Ischemic Heart Disease and Vascular Risk Factors Are Associated With Accelerated Brain Aging

Background Ischemic heart disease (IHD) has been linked with poor brain outcomes. The brain magnetic resonance imaging–derived difference between predicted brain age and actual chronological age (brain-age delta in years, positive for accelerated brain aging) may serve as an effective means of communicating brain health to patients to promote healthier lifestyles. Objectives The authors investigated the impact of prevalent IHD on brain aging, potential underlying mechanisms, and its relationship with dementia risk, vascular risk factors, cardiovascular structure, and function. Methods Brain age was estimated in subjects with prevalent IHD (n = 1,341) using a Bayesian ridge regression model with 25 structural (volumetric) brain magnetic resonance imaging features and built using UK Biobank participants with no prevalent IHD (n = 35,237). Results Prevalent IHD was linked to significantly accelerated brain aging (P < 0.001) that was not fully mediated by microvascular injury. Brain aging (positive brain-age delta) was associated with increased risk of dementia (OR: 1.13 [95% CI: 1.04-1.22]; P = 0.002), vascular risk factors (such as diabetes), and high adiposity. In the absence of IHD, brain aging was also associated with cardiovascular structural and functional changes typically observed in aging hearts. However, such alterations were not linked with risk of dementia. Conclusions Prevalent IHD and coexisting vascular risk factors are associated with accelerated brain aging and risk of dementia. Positive brain-age delta representing accelerated brain aging may serve as an effective communication tool to show the impact of modifiable risk factors and disease supporting preventative strategies.

A lthough the risk of cognitive decline is primarily driven by age, vascular risk factors and coexisting diseases may augment the risk of age-related brain deterioration. 1,2 Ischemic heart disease (IHD) has been associated with poor brain health independently of the effect of aging itself. 3,4 Several pathways involved in heart-brain crosstalk have been proposed, albeit the precise pathophysiological mechanisms by which IHD may cause cognitive deterioration remain to be elucidated. 5 Cardiac ischemia has been linked with macro-and microstructural brain abnormalities in both white and grey matter even before the onset of clinical symptoms of dementia. 6,7 Detecting and quantifying subtle deviations from age-related brain changes may enable identification of subjects at greater risk for developing cognitive decline to whom preventive strategies and targeted treatments should be directed.
Recent neuroimaging studies have introduced modelling approaches to estimate individual brain age based on brain magnetic resonance imaging (MRI) features. The difference between estimated brain age and actual age (delta) reflects deviations from normal aging and may serve as a biomarker for cognitive deterioration. 8 An older-appearing brain indicating accelerated brain aging (positive delta) has been linked with certain neurological conditions and lifestyle factors in cognitively healthy cohorts. 9,10 Brain age may represent an effective means of communicating the risk of brain deterioration to patients, just as vascular age can be used to express cardiovascular disease risk. 11 It could increase risk awareness and likely promote healthier lifestyles and preventive actions. 12 We have investigated the impact of prevalent IHD on brain aging and the risk of dementia, possible mechanisms underlying the association, and the role of vascular risk factors, cardiovascular structure, and function using UKB (United Kingdom Biobank) data.
No previous studies have used this approach to evaluate brain health in large IHD cohorts without pre-existing neurological conditions. We hypothesize that IHD may be associated with accelerated brain aging, reflecting structural brain changes predisposing to cognitive decline or dementia. The estimated brain age can thus be used to express the individual risk level and motivate IHD patients to improve their health behaviors.
SP4, Siemens Healthcare) in accordance with a publicly accessible protocol. 15 In brief, imaging acquisition included 6 different sequences, covering structural, diffusion, and functional imaging for a CMR images were acquired using 1.5-T scanners (MAGNETOM Aera, Syngo Platform VD13A, Siemens Healthcare) according to a predefined acquisition protocol. 16 In brief, the cardiac assessment includes a combination of long-axis cines and a complete shortaxis stack covering both the left ventricle (LV) and right ventricle (RV) acquired using balanced steadystate free precession sequences.
BRAIN MRI FEATURE EXTRACTION. Brain IDPs were extracted from brain MRI and accessible through the UKB showcase (Supplemental Table 2). We selected 25 volumetric features extracted from structural MRIs as they were identified among the top informative predictors to model brain age. 17, 18 The total volume of white matter hyperintensities (WMHs) (not used to model brain age) was instead used for the post hoc mediation analysis. A detailed description of processing data is provided in the Supplemental Methods.
CMR PARAMETERS. Conventional measures of LV and RV structure and function were extracted using an automated pipeline with inbuilt quality control, previously developed, and validated in a large sample of the UKB. 19 Specifically, we analyzed the following parameters: LV and RV volumes in end-diastole and end-systole; LV and RV stroke volumes; LV mass; and left ventricular ejection fraction.
Additional cardiac indices able to capture changes in myocardial function in relation to structural chamber remodeling were analyzed, including LV mass-to-volume ratio, and LV global function index.
Mass-to-volume ratio was calculated as LV mass divided by LV end-diastolic volume. LV global

Post-hoc association analyses
Subjects with brain IDPs (n = 39,125) Excluding subjects with prevalent brain diseases and IHD (n = 3,888) Association of brain-age delta with CMR and vascular risk factors (n = 4,050) Association of brain-age delta with dementia (n = 8,389)  Table 3). Briefly, to assess whether IHD was associated with faster brain aging, whether it, in turn, was linked with dementia risk, and to investigate potential mechanistic links and risk factors involved in the association, we used a staged approach, which can be summarized as follows (Supplemental Figure 1).  Values are mean (AE SD) when continuous or number (%) when categorical. Bold indicates differences between the 2 groups are statistically significant (P < 0.05).
IHD ¼ ischemic heart disease. Values are median (IQR) where absolute skew is $0.9, n (%), or mean AE SD. The data shown here are from the subjects with values that were available. Bold indicates differences between the 2 groups are statistically significant (P < 0.05).
of variance in the predicted brain age explained by the used features in the model. The brain aging model was first built based on participants with non-IHD by splitting the data into 2 subsets (training set, 80%; testing set, 20%). After removing the age-dependency bias (as described in the Supplemental Methods), we calculated brain-age delta by subtracting the chronological age from the predicted brain age in the test data sets, with positive values (expressed in years) indicating accelerated brain aging. Pearson correlation was also calculated between actual age and brain-age delta, before and after bias correction.
Next, brain age was estimated on IHD subjects using the previously trained model to assess the deviation of the brain-age delta from the reference (non-IHD) population. IHD and non-IHD groups were thus compared in terms of model performance and difference in mean brain-age delta, the latter of which was considered as a measure of apparent brain aging. 8 Brain age and risk of dementia. Logistic regression was used to analyze the relationship between brain-age delta and incident dementia (Supplemental Table 4).
The model was adjusted for age, sex, and education level. As the number of incident events was low in comparison to nonevents, to account for imbalance data potentially affecting the model, we repeated the association analysis after using propensity score matching based on age and sex.
Mediation effects of WMH on IHD and brain age. To study potential mechanistic links, we performed a mediation analysis to test to which extent WMH, a marker of cerebrovascular injury, mediated the impacts of IHD on brain aging (indirect effect). The ordinary least squares regression was used to study the associations (described in terms of effect) between the variables (IHD, WMH as mediator, and brain-age delta). The model was adjusted for age, sex, and education.
Association of brain age with vascular risk factors and imaging parameters. A linear regression model in both IHD and non-IHD (test-set) groups was used to assess the role of vascular risk factors and imaging parameters in advancing brain age. The model was adjusted for the same confounds used for brain aging model plus age. In the analysis, we used unstandardized measures to reveal the effect (beta value) of changing 1 unit in the exposure on the brain-age delta. Specifically, changing the exposures may lead to increasing or decreasing in brain-age delta (estimated in years) based on the direction of the association (positive vs negative beta value).
As we found significant associations between certain imaging parameters and brain aging only in non-IHD subjects, we studied whether these metrics were linked with incident dementia in this group using logistic regression. The model was adjusted for age and sex.

BASELINE CHARACTERISTICS OF THE STUDY
POPULATIONS. People with IHD were on average older than the controls (59.2 AE 6.3 years vs 54.6 AE 7.4 years; P < 0.05), albeit the age range was similar in the 2 cohorts (40 to 70 years) (Supplemental Figure 2).
The diseased subjects were more likely to be males than those without the disease (71.4% vs 46.3%, respectively; P < 0.05) ( Table 1), with a greater burden of risk factors and worse cardiovascular health indices ( Table 2). IHD subjects had significantly lower volumes of most brain structures and increased volume of WMH than those without the disease (Supplemental Table 5).
ESTIMATED BRAIN AGE IN IHD. In non-IHD subjects (test set), the model showed a mean difference between the chronological age and predicted brain age (MAE) of 4.69 years ( Table 3). When applying the model to the IHD group, a greater deviation of predicted brain age from actual age was observed (MAE ¼ 6.96 years) indicating an older-appearing MAE indicates the mean difference between the chronological age and the predicted brain age expressed in years, with higher values indicating older-appearing brain. R 2 represents the proportion of variance in the predicted brain age explained by the used features in the model. In IHD group, a greater deviation of predicted brain age from chronological age was observed compared to non-IHD (test-set) (MAE ¼ 6.96 years for IHD vs MAE ¼ 4.69 years for non-IHD).
MAE ¼ mean absolute error; R 2 ¼ coefficient of determination; other abbreviation as in Table 1. brain. A significant difference in mean brain-age delta values between the 2 cohorts was also found (P < 0.001), confirming that the IHD subjects had significantly higher brain age than those without the disease.
The difference between the 2 groups was further confirmed by R 2 , with a negative value in IHD indicating that the prediction in this group was worse than the mean prediction (Supplemental Figure 3).
Such negative value was mentioned in the scikit-learn documentation and reported in previous brain aging studies. 25 The correlation value between brain-age delta and actual age in both cohorts decreased to close to zero after the estimated brain age was corrected from bias.
This indicates that the correction steps were performed correctly, and the derived brain-age delta was free from age dependency. This was notably observed in the IHD cohort (Supplemental Table 6).

ASSOCIATION OF BRAIN AGE WITH INCIDENT
DEMENTIA. One unit increase in brain-age delta was associated with an increase of 13% in the odds of dementia occurring (odds ratio:  Table 4).

MEDIATION EFFECTS OF WMH ON IHD AND BRAIN
AGE. The association of IHD diagnosis with brain-age delta can be explained as a direct effect, caused by the disease itself, an indirect effect mediated by WHM, and a total effect, which is the result of both ( Figure 2).
An unadjusted model showed that IHD had significant effects on brain-age delta both directly and indirectly (P < 0.001). However, the direct effect of IHD on brain aging was far larger than through WMH as mediator (beta values ¼ 6.41 vs 0.11, respectively).
When adjusted for covariates, the direct effect remained similar in size and significant (beta value ¼ 6.62; P < 0.0001), while the mediation effect of WMH became nonsignificant.

ASSOCIATION OF BRAIN AGE WITH VASCULAR RISK
FACTORS. Increasing brain-age delta was significantly associated with a history of diabetes in both groups. However, such effect was greater in IHD (beta values up to 2.1 years in IHD vs 1.3 years in controls).
In IHD, faster brain aging was also significantly linked with increased adiposity (larger waist-hip ratio and increased body mass index). The greatest association was found with waist-hip ratio, where for 1 unit increase of such value, there was an increase in estimated brain age of up to 8.02 years from normal aging (as indicated by the beta value) ( Table 5).   The beta value is interpreted as the difference in the dependent variable (brain-age delta) for changing 1 U in the independent variable (exposure). Specifically, changing 1 U in the exposure leads to increasing (positive beta value) or decreasing (negative beta value) in brain-age delta. For instance, in non-IHD, diabetes is associated with an increased brain-age delta ¼ 1.3 years, whilst in IHD such increase is near 2.1 years. Bold indicates statistically significant after Bonferroni correction (P < 0.05). a Indicates categorical variables; the other variables are continuous.
Abbreviations as in Tables 1 and 2. risk factors on brain aging in the 2 cohorts is shown in Figure 3.

ASSOCIATION OF BRAIN AGE WITH CMR PARAMETERS.
In non-IHD subjects, increasing brain-age delta was significantly associated with smaller LV and RV volumes, and lower indices of cardiac function, LV mass, aortic distensibility, and TAC values. Similar direction of associations was found in the IHD cohort for all imaging metrics, albeit not statistically significant ( Table 6). None of the CMR metrics were significantly associated with the risk of dementia (Supplemental Table 7).

DISCUSSION
In this study, we estimated the brain age of IHD subjects from the UKB using volumetric brain features extracted from MRI data. We found that subjects with IHD showed, on average, a more advanced brain age than their chronological age, with their brains appearing significantly older than those without the disease. Accelerated brain aging was also linked with increased risk of dementia. WMH, a marker of cerebrovascular injury, did not significantly mediate the effects of IHD on brain aging, suggesting that other mechanisms related to the disease itself were involved in the association. IHD, diabetes, and increased adiposity were associated with even faster brain aging. Overall, these findings suggest that IHDrelated mechanisms and vascular risk factors concur to accelerate biological brain aging and so the risk of developing dementia (Central Illustration).

Non-IHD Cohort IHD
Higher beta-value (in years) indicates faster brain aging. An asterisk (*) indicates that the association is statistically significant (P < 0.05).
Created using BioRender.com. BMI ¼ body mass index; other abbreviations as in Figure 1. given the different approaches used to evaluate the impact on brain structures, they are consistent with existing research and confirm the association between IHD and poorer brain health. 4 There is evidence to support faster cognitive deterioration in the years following the diagnosis of IHD than before, suggesting possible disease-related mechanisms implicated in the association. 26 We observed that IHD subjects had early structural brain changes in the form of brain atrophy and increased WMHs than their peers. WMHs reflect underlying cerebral small vessel disease, a form of vascular dysfunction which may occur both in the brain and the heart, leading to different clinical manifestations, including ischemia, cognitive dysfunction, and dementia. 27,28 However, we observed that the vascular contributions to brain aging in IHD, as expressed by WMHs, were nearly insignificant. Instead, it was mainly a direct effect of the disease to drive the association. These results confirm the strong association between IHD and brain deterioration, and that such relationship is not fully explained by WMHs. 6,7 Our findings suggest that IHD contributes to brain aging through several disease-related mechanisms that this marker of cerebrovascular injury could not capture. Among the mechanisms involved are direct ischemic brain injury, blood-brain barrier alterations, and various biological pathways, including oxidative stress, immune responses, and endothelial dysfunction. 5 Future studies using more sensitive measures of brain microstructural changes are needed to

CENTRAL ILLUSTRATION Ischemic Heart Disease and Vascular Risk Factors Accelerate Brain Aging
High adiposity Similarly, we observed that increased adiposity and diabetes were associated with significantly faster brain aging in presence of IHD. This indicates that the impact of such factors on brain structures is amplified when there is a coexisting diagnosis of IHD with a synergistic effect of vascular and ischemic-related processes on disease progression.
Diabetes played a significant role in brain aging even in the absence of IHD, confirming that this condition is one of the strongest risk factors for dementia. 29 We also found a very strong association between indices of adiposity and brain aging in IHD.
A higher waist-to-hip ratio, a measure of abdominal However, future prospective trials showing the effectiveness of using brain age for communicating risk information are needed before recommending such an approach in clinical practice.
STUDY LIMITATIONS. We used volumetric brain features to estimate brain age as they were previously identified among the top informative predictors.
However, including multiple imaging modalities could lead to better model performance.
The number and type of brain IDPs used in the model can affect the estimated value of brain-age delta. However, our purpose was to use this parameter to show the differences in brain aging rate between IHD vs non-IHD rather than propose a quantification of aging itself.
Additional factors, including genetics and early brain developments, can likely influence brain structures. Future studies based on longitudinal imaging data could help determine such factors' role in the individual change trajectories of brain aging.
The average actual age of IHD and non-IHD differed significantly. However, we applied an agebias correction method that eliminated the effects of age from the estimated brain age delta.
We assessed the effect of certain previously validated exposures linked to brain health. Additional covariates, including biological, psychological, and behavioral parameters, were not evaluated. That was done to avoid model overfitting, potentially decreasing the sensitivity to real effects. Differences in sample size, age range, and methods used to select covariates and brain features might explain some discrepancies between the current and previous brain aging results on UKB.
Although the direction of associations between exposures and brain-age delta might indicate the potential impact of the exposure itself on brain aging (accelerating vs decelerating), these results do not allow for causal inference. We cannot infer temporal relationships between brain aging and exposures as this is a cross-sectional study. Future longitudinal investigations using serial brain and CMR data linked with clinical outcomes might evaluate the effect of having an older appearing brain in the long term.

CONCLUSIONS
Prevalent IHD is associated with accelerated brain aging and increased risk of dementia, that is not fully explained by microvascular injury. Besides shared vascular risk factors, additional disease-related mechanisms might contribute to abnormal aging.
Using brain age to communicate the risk levels for cognitive deterioration may increase patients' awareness and improve their adherence to therapies and lifestyle changes.