Title:
Amyloid-β and Vascular Risk Shape Brain-Aging Pathways to Cognition in Cognitively Normal Adults: A Harmonized Multi-Cohort Analysis

Introduction:
In cognitively normal adults, regional brain structure is shaped by both amyloid-β (Aβ) pathology and modifiable vascular risk factors. Prior studies have linked Aβ and vascular burden separately to brain atrophy and worse cognitive performance, but it remains unclear whether these effects converge on shared brain-aging pathways. In particular, vascular risk may influence cognition through multiple structural levels: raw regional brain volumes, deviation from age- and sex-referenced normative structure, and global accelerated brain aging. Large, harmonized multi-cohort imaging analyses are therefore needed to determine whether vascular risk affects cognition through brain-aging markers that overlap with Aβ-related structural vulnerability.

Methods:
Cross-sectional baseline data were used from cognitively normal adults in HABS, ADNI, MCSA, and EPINETTE, with approximately 3,000 participants aged 43–85 years. Only the first visit was retained for each participant. All participants had T1-weighted MRI, Aβ status, and MMSE.

T1 MRI was processed with the ANTs cortical thickness and volumetric pipeline. Regional ROI volumes were harmonized across cohorts using CovBat, preserving age and sex effects. Aβ status was determined per cohort using cohort-specific cut-points from PET, CSF, or plasma and encoded as binary Aβ− or Aβ+. Vascular exposures were binary indicators from clinical records or self-report, including hypertension, diabetes, stroke, ever smoking, and obesity.

Three complementary brain-aging representations were examined. First, raw regional MRI features were used to characterize Aβ-related structural patterns. Sex-stratified generalized additive models modeled nonlinear age effects and tested ROI-specific Aβ × age interactions, with FDR correction across ROIs. Second, reference centiles were derived from Aβ− participants by ranking GAM-adjusted residual ROI volumes after accounting for age, sex, education, and whole brain volume. Lower centiles indicated smaller-than-reference regional volumes. Third, brain age gap (BAG) captured accelerated global brain aging at the subject level, with higher BAG indicating older-appearing brain structure relative to chronological age.

Two-equation OLS mediation models with bootstrap inference were run separately using BAG and each regional ROI reference-centile score as the mediator. The Aβ-moderated extension tested exposure × Aβ and mediator × Aβ interactions to estimate indirect effects in Aβ− and Aβ+ participants.

Results:
Harmonization substantially reduced site and cohort effects while preserving biological age and sex effects. After harmonization, ROI volume distributions showed greater overlap across HABS, ADNI, MCSA, and EPINETTE, supporting pooled cross-cohort analysis.

Brain-aging markers differed by Aβ status and sex. Aβ+ participants showed higher BAG than Aβ− participants, indicating greater global brain-aging burden. Reference-centile ROI score distributions also differed by Aβ status, suggesting region-specific Aβ-related brain-aging effects beyond global BAG.

Raw MRI feature analyses identified Aβ-related regional structural patterns. Sex-stratified models showed Aβ-associated lower regional volumes, particularly in medial and lateral temporal regions. Significant age-dependent Aβ effects were observed in bilateral putamen and right inferior parietal volume, with stronger effects in females.

BAG mediated the association between vascular exposures and cognition. Joint-exposure mediation models showed that all five vascular exposures were associated with accelerated brain aging, which in turn was associated with lower MMSE. Stroke and ever smoking showed the largest BAG-mediated effects. Aβ-moderated mediation models indicated that Aβ pathology amplified selected vascular risk–brain aging–cognition pathways, primarily by increasing the cognitive cost of a given level of brain aging.

Reference-centile mediation analyses revealed regional brain-aging pathways linking vascular risk to cognition. Vascular risk–cognition mediation pathways converged on temporal cortical regions, including entorhinal, parahippocampal, and middle temporal areas. Global and white matter volumes also mediated associations for selected exposures such as diabetes and stroke. Aβ amplified several centile-mediated pathways, suggesting that amyloid pathology may strengthen both the effect of vascular risk on structural deviation and the effect of structural deviation on cognition.

A joint-exposure mediation network summarized significant pathways linking vascular exposures to MMSE through global and regional brain-aging mediators. Vascular exposures were connected to BAG and multiple ROI aging mediators, and Aβ acted as a moderator of selected indirect pathways.

Conclusions:
In cognitively normal adults, vascular risk factors influence cognition through brain-aging structural pathways that include both accelerated global brain aging and region-specific deviation from normative Aβ− structural aging. Aβ pathology is associated with lower temporal regional volumes and amplifies selected vascular risk–brain aging–cognition pathways. The convergence of amyloid- and vascular-related effects in medial and lateral temporal cortex suggests a shared structural vulnerability zone. These findings support the value of combining raw MRI features, reference-centile scores, and BAG to identify early brain-aging pathways through which modifiable vascular risks may affect cognition before clinical impairment emerges.

References:
Chen AA et al. Harmonization of neuroimaging data using CovBat. NeuroImage, 2022.
Bethlehem RAI et al. Brain charts for the human lifespan. Nature, 2022.
Smith SM et al. Estimation of brain age delta from brain imaging. NeuroImage, 2019.