APOE ε4 sharpened the link between left atrial remodeling and later brain change
APOE ε4 appears to change how subclinical cardiac remodeling maps onto brain structure with age, and the clearest signal in this preprint came from the left atrium. In Framingham Heart Study Offspring participants, antecedent left atrial remodeling was tied to later lateral ventricular expansion in an APOE ε4-dependent way, with the interaction surviving false-discovery-rate correction across the longitudinal MRI analyses. The result is biologically plausible: chronic atrial remodeling can reflect cumulative hemodynamic stress, impaired diastolic filling, and a vascular milieu that may influence downstream brain vulnerability over years rather than months.
The study also points to a narrower pattern than a broad “cardiac aging equals brain aging” story. Across 16 longitudinal tests, this was the only result that remained significant after FDR correction. The effect size was small, as expected for structural epidemiology, but the statistical signal was strong enough to persist in models that incorporated participant-specific MRI-time slopes. When intracranial volume by MRI time entered the model, the association weakened, which argues for caution in treating any single morphometric finding as a definitive disease pathway.
Ventricular expansion carried the strongest longitudinal signal
Investigators related echocardiographic remodeling measured in Framingham examinations 4 through 6 to repeated brain MRI, using 1,305 participants and 3,787 MRI observations. The key longitudinal association centered on how APOE ε4 modified the relationship between antecedent left atrial remodeling and subsequent lateral ventricular expansion, with a beta of −0.00858, a P value of 1.45×10⁻⁴, and q=0.0023. The direction of the interaction indicates that APOE ε4 status altered the slope of the atrial-brain association rather than simply shifting baseline brain size.
One sentence captures the clinical tension: the strongest signal involved a chamber change that often advances silently for years. Left atrial enlargement can sit below the threshold of routine bedside concern, yet here it tracked with later ventricular enlargement in a genotype-dependent fashion. That does not prove causality, but it does suggest that cardiac structure may carry information about brain aging long before overt stroke, heart failure, or cognitive syndromes appear.
Regional analyses were more complicated. Four ventricular cardiac-by-APOE ε4 interactions met within-family FDR criteria, but permutation testing showed that the number was compatible with chance, with permutation P=0.70. In other words, the regional ventricular findings did not hold up as a coherent family of discoveries once the analysis was stress-tested against random expectation.
Occipital cortex variability and the limits of regional inference
Beyond the ventricular findings, maximum left atrial volume index showed an APOE ε4-dependent association with left middle-occipital cortical-thickness variability. The whole-brain result reached q=0.017, and permutation testing gave P=0.041. This association persisted after ventricular adjustment, which makes it less likely to be a simple byproduct of global brain size or ventricular enlargement alone.
The occipital cortex result is intriguing because it does not immediately map onto the most commonly discussed brain networks in aging or Alzheimer’s disease. Visual association cortex is not the first place most clinicians would look for a cardiac remodeling signature. Yet the brain does not age uniformly, and region-specific vulnerability may reflect vascular supply, white matter connectivity, or genotype-mediated susceptibility to structural change. Even so, a single regional association in a broad screen should be treated as hypothesis-generating, not practice-changing.
Biomarkers of neurodegeneration did not explain the cardiac findings
The study also tested whether APOE ε4-dependent cardiac associations were mirrored by plasma biomarkers of neurodegeneration and glial or axonal injury. There were no FDR-significant interactions with plasma Aβ42/40, p-tau181, GFAP, or NfL. Adjusting for these biomarkers altered the regional interactions by no more than 3.2%, which argues against a major mediating role for the sampled blood markers in these specific associations.
That absence matters because it narrows the interpretation. If the atrial-brain connection were strongly driven by overt amyloid, tau, astroglial activation, or neuroaxonal injury in plasma, the biomarker analyses would have been more likely to move the estimates. Instead, the data suggest that the cardiac-genotype relationship may operate through other routes, including long-term vascular burden, microembolic risk not captured here, or shared susceptibility to remodeling across organ systems.
A genetics-first pattern, but not a final mechanism
Exploratory pathway work using Allen Human Brain Atlas gene-expression data found cholesterol-biosynthesis enrichment in the right superior frontal cortex, the region implicated by the left ventricular end-systolic volume interaction. Even there, convergent pathway evidence was lacking. The molecular signal therefore remains suggestive rather than persuasive, and it cannot be used to claim a single mechanistic axis connecting atrial remodeling, ventricular change, and APOE ε4 status.
APOE ε4 remains relevant because it is not only a dementia-risk allele; it also marks altered lipid handling, vascular biology, and possibly brain resilience. Those pleiotropic effects provide a credible framework for why the same degree of cardiac remodeling might map differently onto brain morphology depending on genotype. In carriers, a cardiovascular phenotype that looks subclinical on routine imaging may reflect a biologic environment less able to buffer downstream structural consequences.
What the Framingham data support, and what they do not
The analysis drew on a well-characterized community cohort, controlled-access Framingham and dbGaP resources, and repeated MRI measurements, which strengthens internal consistency. It also used prespecified interaction testing, false-discovery-rate correction, and permutation testing for the regional screen. Those design features make the positive findings more credible than a simple one-off association scan.
Even so, the preprint leaves important uncertainty. The strongest association was longitudinal, but it was still observational. The study cannot separate atrial remodeling from the long chain of clinical and subclinical processes that accompany aging, blood pressure exposure, metabolic stress, and vascular injury. The data also do not establish whether intervening on atrial remodeling would alter brain trajectories, or whether the atrial signal is simply a marker of broader systemic aging.
Exploratory analysis also found that HbA1c related to left ventricular end-diastolic volume differently by APOE ε4 status, but that result sits outside the main longitudinal story and should be read as preliminary. The more defensible conclusion is narrower: in this Framingham analysis, APOE ε4 modified how left atrial remodeling related to later brain structural change, and the most convincing signal involved ventricular expansion rather than a wide set of regional cortical or fluid biomarkers. Replication with prespecified models is needed before this pattern can be translated into risk stratification or mechanistic inference.