When Alzheimer's pathology meets cardiometabolic risk: intrinsic subcortical-cortical connectivity signatures of retroactive interference in aging

BACKGROUND: Retroactive interference (RI), the disruption of past memory by new information, has been proposed as a sensitive marker of early vulnerability in aging and Alzheimer’s disease (AD). Yet, its neural bases and links to cardiometabolic risk remain largely unknown. Here, we investigated how well-established AD biomarkers interact with cardiometabolic factors to shape intrinsic subcortical–cortical functional connectivity and RI performance in aging. METHODS: Seventy-seven cognitively normal older adults (52–79 years, 47 women) underwent cognitive, neuroimaging, and AD biomarker assessments, including an interference task, resting-state fMRI, global cortical amyloid-β (Aβ) load, plasma pTau-181, and serum GFAP and NfL. A latent cardiometabolic factor was derived from fasting glucose and triglycerides, while systolic blood pressure served as an independent vascular index. Seed-to-cortex functional connectivity was analyzed using regression models testing how RI–related connectivity was modulated by the combined effects of AD biomarkers and metabolic or vascular risk factors. RESULTS: Increased connectivity within hippocampo–thalamic–cortical networks was associated with greater RI among individuals with higher global cortical Aβ load and elevated metabolic risk. Under elevated pTau-181 levels, metabolic and vascular status jointly shaped the relationship between functional connectivity and RI, extending to broader cerebral networks including basal forebrain cholinergic nuclei. Similarly, in the context of elevated GFAP levels, cardiometabolic factors modulated connectivity–RI associations: under higher metabolic risk, enhanced subcortical–fronto–temporal connectivity predicted greater RI, whereas under high vascular stress, stronger subcortical–fronto–parietal connectivity predicted lower RI. Finally, elevated NfL, combined with vascular risk was associated with enhanced subcortical–frontal connectivity and lower RI. Crucially, these RI–related connectivity signatures were not evident for proactive interference, while some of the signatures overlapped with broader associative memory measures, suggesting that RI primarily reflects interference-specific processes, with neural substrates that may overlap partially with broader memory networks in cognitively normal older adults. CONCLUSIONS: Overall, our results indicate that RI–related functional connectivity patterns in cognitively normal older adults are selectively shaped by the combined influence of AD biomarker levels and cardiometabolic risk, providing a framework for early identification of subtle cognitive vulnerabilities and the development of interventions to support memory function in aging.