APOE ε4 and peripheral metabolic network architecture in older adults

Background Whether apolipoprotein E ( APOE ) ε4 status shapes the organization of modifiable peripheral metabolic risk pathways for dementia prevention remains unclear. Most prior studies have focused on individual APOE -by-biomarker interactions rather than the broader conditional dependency structure among metabolic variables. Objective We aimed to examine whether APOE ε4 is conditionally connected to metabolic factors and whether network architecture differs by genotype. Methods In 2454 participants from the Korean Longitudinal Study on Cognitive Aging and Dementia with complete data on 13 variables, conditional dependency networks were estimated using graphical least absolute shrinkage and selection operator models and mixed graphical models. Network Comparison Tests were used to compare 12-node metabolic networks between ε4 carriers and non-carriers. Cox proportional hazards models evaluated whether hub biomarkers predicted incident cognitive impairment over a median follow-up of 7.4 years and whether associations differed by APOE genotype. Results APOE ε4 showed negligible conditional connectivity, with near-zero partial correlations in GLASSO (maximum |r| = 0.0114) and zero edge weights in mixed graphical models. The 12-node metabolic network did not differ significantly between carriers and non-carriers in either network structure (M = 0.1671, p = 0.45) or global strength (S = 2.3662, p = 0.46). Among 1922 baseline cognitively normal participants, 504 developed cognitive impairment during follow-up. Neither the APOE × homocysteine interaction (p = 0.921) nor the APOE × creatinine interaction (p = 0.402) was significant. Conclusions APOE ε4 was conditionally independent of the peripheral metabolic variables examined, with no major genotype-specific differences in network architecture. These findings highlight the importance of metabolic dementia prevention regardless of genotype, although interventional confirmation is warranted.