A Multi-Axis Framework for Late-Life Alzheimer’s Disease Interpretation

Late-life Alzheimer's disease (AD) is increasingly defined by biomarkers, yet in adults aged ≥65 years the relationship between amyloid positivity and near-term cognitive decline is often discordant. Amyloid PET robustly detects fibrillar plaque burden, but it incompletely captures dynamic and potentially neurotoxic amyloid processes, particularly soluble assemblies and oligomer-related "activity." This review rethinks the late-life AD spectrum by integrating four clinical lenses that frequently drive real-world interpretive uncertainty: (1) amyloid PET positivity as a measure of fibrillar plaque presence and magnitude; (2) plasma amyloid-β oligomerization tendency measured by the multimer detection system (MDS-OAβ) as an activity-oriented (i.e., a dynamic readout hypothesized to reflect ongoing processes rather than cumulative lesion burden), process-linked readout that may decouple from plaque burden; (3) postoperative delirium (POD) as a time-anchored stress-test phenotype revealing vulnerability and reduced resilience under systemic insults; and (4) drug-linked biomarker trajectories, contrasting rapid plaque removal by anti-amyloid monoclonal antibodies with observational signals raising the hypothesis that Ginkgo biloba may be associated with oligomer-related biology and, in some contexts, differences in longitudinal amyloid accumulation trajectories in the absence of observed immediate plaque reduction. We propose a pragmatic multi-axis framework-plaque burden, amyloid activity, downstream engagement, and vulnerability/resilience-to contextualize late-life discordances such as PET positivity without decline, PET negativity with elevated MDS-OAβ, delirium-associated decompensation, and clinical change without rapid PET decline. This synthesis highlights testable predictions and prioritizes longitudinal, multi-marker studies to determine whether activity-oriented biomarkers and stress phenotypes can refine late-life risk stratification beyond plaque-centered models.