🧠 neurology

Cardiovascular risk and hippocampal-cognitive coupling in Alzheimer's disease

This study demonstrates that while the standard Framingham Risk Score predicts hippocampal atrophy and cognitive decline largely due to age-related variance, an age-adjusted cardiovascular risk measure (CVRmimic) fails to show these associations, suggesting that age-adjusted metrics are necessary to isolate true cardiovascular effects from aging in Alzheimer's disease.

Original authors: Fernandez-Lozano, S., Villeneuve, S., Collins, D. L., for the Alzheimer's Disease Neuroimaging Initiative,

Published 2026-06-03

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Fernandez-Lozano, S., Villeneuve, S., Collins, D. L., for the Alzheimer's Disease Neuroimaging Initiative,

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

The Big Question: Is it the Heart or the Clock?

Imagine you are trying to figure out why an old car is breaking down. You have a "Risk Score" that predicts how likely the car is to have a major engine failure in the next 10 years. This score is very high for older cars.

The big question this study asks is: Is the car breaking down because it's old (the clock), or because it has bad maintenance habits (the heart/vascular risk)?

In the world of Alzheimer's research, scientists often use a tool called the Framingham Risk Score (FRS) to measure cardiovascular health. They noticed that people with higher scores tend to have smaller memory centers in their brains (the hippocampus) and worse thinking skills. But because the FRS is heavily weighted by age (older people get higher scores automatically), it's hard to tell if the brain damage is caused by the heart issues or just by getting older.

The Experiment: Two Different Rulers

To solve this puzzle, the researchers compared two different ways of measuring heart risk in 923 people with early Alzheimer's disease:

The Standard Ruler (FRS): This is the usual score. It includes age, blood pressure, smoking, etc. It's like a thermometer that gets hotter just because the room is older.
The "Age-Stripped" Ruler (CVRmimic): The researchers used a special statistical trick (called a MIMIC model) to take the "age" part out of the score. This new score measures heart risk independent of how old you are. It's like measuring the engine's condition while ignoring the car's mileage.

What They Found

The results were surprising and clear:

1. The "Old Car" Effect (FRS)
When they used the standard FRS, it looked like heart risk was causing the memory centers of the brain to shrink faster.

The Analogy: It's like saying, "Older cars break down faster." This is true, but it's mostly because they are old, not necessarily because they have bad oil. The study found that the FRS was mostly just picking up on the fact that the participants were aging.

2. The "Real Heart" Effect (CVRmimic)
When they used the new "age-stripped" score, the link disappeared.

The Analogy: Once they removed the "age" factor, the "bad maintenance" score didn't predict that the engine would shrink any faster. This suggests that in people with Alzheimer's, the standard heart score isn't actually driving the brain shrinkage; it's just a shadow of aging.

3. The Brain-Cognition Connection (The Coupling)
Here is the most interesting part: Even though the "age-stripped" heart score didn't cause the brain to shrink, the link between the brain's size and thinking skills remained strong.

The Analogy: Think of the brain (hippocampus) as a library and thinking skills as the books you can read. The study found that no matter how you measure heart risk, if the library gets smaller (atrophy), the ability to read books (cognition) goes down. The heart risk didn't cause the library to shrink, but the size of the library still perfectly predicts how well you can read.

4. The "Reserve" Effect
In a small group of people who didn't have Alzheimer's pathology (no amyloid plaques), the "age-stripped" heart score actually showed a different pattern. It seemed to act like a "reserve."

The Analogy: For people without the disease, having good heart health (even after removing age) seemed to help them keep their thinking skills sharp, even if their brain structure was changing. It's like having a spare tire that helps you keep driving smoothly even if the road gets bumpy.

The Takeaway

The paper concludes that when we see a link between heart risk and brain shrinkage in elderly people with Alzheimer's, we are mostly seeing the effect of aging, not the effect of heart disease.

The Standard Score (FRS) is like a shadow cast by the sun (age). It looks big and scary, but it's just the sun.
The New Score (CVRmimic) tries to measure the object casting the shadow. It found that the object (heart risk) isn't actually pushing the brain to shrink faster in these patients.
However, the size of the brain (hippocampus) remains a very reliable "fuel gauge" for how well a person is thinking, regardless of how we measure their heart health.

In short: The study suggests that for people with Alzheimer's, the brain shrinking is driven by the disease and aging, not directly by cardiovascular risk factors once you account for age. But, the brain's structure remains a perfect predictor of how well a person thinks.

Technical Summary: Cardiovascular Risk and Hippocampal-Cognitive Coupling in Alzheimer's Disease

Problem Statement
The Framingham Risk Score (FRS) is widely used to index cardiovascular risk (CVR) and has been associated with reduced brain volumes and accelerated cognitive decline in aging cohorts. However, the FRS is heavily weighted by age (log-hazard coefficients of 3.06 for men and 2.33 for women), making age the dominant predictor within the composite score. In elderly samples, this creates a confounding scenario where age acts as a common cause for both the predictor (FRS) and the outcome (neurodegeneration). Standard regression controls for age are insufficient because age is an internal component of the score itself, not merely an external covariate. Consequently, it remains unclear whether FRS associations with brain atrophy reflect true cardiovascular pathophysiology or are artifacts of age-weighting. This study aims to disentangle these effects by comparing the standard FRS against an age-adjusted latent measure of cardiovascular risk.

Methodology
The study utilized data from 923 amyloid-positive participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI), including individuals with cognitively unimpaired (CU), mild cognitive impairment (MCI), and Alzheimer's disease (AD) diagnoses. Analyses were sex-stratified.

Risk Assessment:
- FRS: Derived from the ADSP-PHC dataset using the Framingham 2008 general CVD risk algorithm.
- CVRmimic: A novel age-adjusted measure derived via a Multiple Indicators Multiple Causes (MIMIC) structural equation model. In this model, age is treated as an exogenous cause of a latent cardiovascular risk factor. The residual term ( $\zeta$ ), representing cardiovascular risk independent of age, was extracted as the CVRmimic score. This score is orthogonal to age by construction ( $r \approx 0$ ). Five indicators were used: systolic blood pressure, hypertension history, fasting glucose, total cholesterol, and serum creatinine.
Neuroimaging and Cognitive Metrics:
- HVR: The Hippocampal-to-Ventricle Ratio (HVR) was calculated from T1-weighted MRI volumes. HVR z-scores were derived using Generalized Additive Models for Location, Scale, and Shape (GAMLSS) trained on UK Biobank data to normalize for age, sex, education, and head size.
- Cognition: ADSP-PHC harmonized composite scores for memory, language, and executive function were used. Measurement error was incorporated by fixing residual variances to the squared standard errors of the composites.
- Time Metric: Estimated Disease Time (EDT) was used instead of calendar time to align participants on a common disease progression timeline relative to AD dementia diagnosis.
Statistical Frameworks:
- Linear Mixed-Effects (LME) Models: Tested for a three-way interaction (Years $\times$ CVR $\times$ HVR) to determine if cardiovascular risk moderates the coupling between hippocampal integrity and cognitive performance over time.
- Latent Growth Curve Mediation (LGCM) Models: Tested a mediation pathway where CVR predicts the slope of HVR decline (a-path), which in turn predicts the slope of cognitive decline (b-path). This assessed whether CVR drives cognitive decline through accelerated hippocampal atrophy.
- Validation: A separate sample of amyloid-negative, cognitively unimpaired participants was used to validate the normative transfer of z-scores and test for reverse patterns indicative of cognitive reserve.

Key Results

FRS vs. CVRmimic on Hippocampal Atrophy: In LGCM models, the standard FRS significantly predicted the rate of hippocampal atrophy (a-path) in all six models (three cognitive domains $\times$ two sexes). In contrast, the age-adjusted CVRmimic showed no significant effect on hippocampal atrophy in any of the six models.
Hippocampal-Cognitive Coupling: The coupling between hippocampal decline and cognitive decline (b-path) remained significant in four of six models for both FRS and CVRmimic. This indicates that the relationship between structural brain integrity and cognitive performance persists regardless of how cardiovascular risk is measured.
Mediation Effects: Indirect effects (CVR $\to$ HVR slope $\to$ Cognitive slope) were significant for FRS in four of six models but were null for CVRmimic in all six models.
LME Interactions: The three-way interaction (Years $\times$ CVR $\times$ HVR) was significant for FRS across all six models (surviving FDR correction). For CVRmimic, only one interaction (Female Executive Function) reached uncorrected significance, and none survived FDR.
Amyloid-Negative Controls: In the amyloid-negative, cognitively unimpaired control sample, the pattern reversed: CVRmimic showed a significant interaction while FRS did not, suggesting a potential cognitive reserve effect masked by amyloid-driven atrophy in the AD+ sample.

Significance and Claims
The paper concludes that the apparent effects of cardiovascular risk on hippocampal-cognitive decline in elderly cohorts are largely driven by age-weighting inherent in the FRS, rather than by independent cardiovascular pathophysiology. By removing the age component via the MIMIC framework, the study demonstrates that:

Age-Confounding: Previous reports linking FRS to brain atrophy likely conflate age-driven variance with cardiovascular effects.
Decoupling of Mechanisms: While age-adjusted cardiovascular risk does not appear to accelerate hippocampal atrophy (null a-path), the coupling between hippocampal structure and cognition remains robust.
Reserve Mechanism: The reverse pattern observed in amyloid-negative controls suggests that cardiovascular burden may modulate the brain-cognition relationship via a reserve-like mechanism (moderation without mediation) rather than by directly accelerating neurodegeneration.
Methodological Utility: The MIMIC-based age-adjustment framework is presented as a generalizable tool for isolating the effects of composite scores where a constituent variable (like age) independently predicts the outcome, offering a more precise approach for designing intervention trials and interpreting risk in aging populations.

The Big Question: Is it the Heart or the Clock?

The Experiment: Two Different Rulers

What They Found

The Takeaway

Technical Summary: Cardiovascular Risk and Hippocampal-Cognitive Coupling in Alzheimer's Disease

More like this