Abstract
Metabolic dysregulation is a hallmark of Alzheimer's disease (AD), yet the temporal nature of metabolite-phenotype associations remains poorly understood. We systematically evaluated 506 serum metabolites across 4,063 longitudinal samples from 1,430 participants in the Alzheimer's Disease Neuroimaging Initiative (ADNI), applying cross-sectional single-timepoint analyses, multi-timepoint meta-analysis, and time-interaction analysis. Across 15 AD-related phenotypes, we identified 311 metabolites to be significantly associated with disease. Of those, 281 emerged from the multi-timepoint meta-analysis, 243 (216 overlapping/27 additional) from cross-sectional analyses, and 19 (16 overlapping/3 additional) metabolites that showed a significant evolution of their association with AD over time. In total, 128 metabolites (41%) showed persistent associations over time, providing evidence for chronic and systemic metabolic dysregulation in the disease. This, together with the comparably small number of metabolites showing evolving changes, suggests that many metabolic alterations in AD do not change substantially anymore once they manifested. Our findings confirm impaired fatty acid and energy metabolism, disrupted neurotransmitter systems, and oxidative stress as key metabolic features of AD. We demonstrate broad replication of the reported metabolite associations in prior studies and an independent lipidomics dataset in ADNI. In summary, this work expands previous metabolomics studies in AD and provides novel leads regarding timing and persistence of metabolic alterations across the disease trajectory.