Association Between Predicted 10-Year Atherosclerotic Cardiovascular Disease Risk and Magnetic Resonance Imaging-Defined Covert Brain Infarction
Article information
Dear Sir:
Covert brain infarction (CBI) is a clinically silent cerebrovascular
lesion that doubles the risk of future stroke and is strongly associated with cognitive decline [1-3], yet its detection depends almost exclusively on magnetic resonance imaging (MRI), which limits its feasibility for large-scale screening in asymptomatic populations. As a result, most individuals with covert cerebral ischemic injury remain unidentified until overt neurological events occur. The 10-year atherosclerotic cardiovascular disease (ASCVD) risk score provides an integrated assessment of global vascular risk and is widely used in routine clinical practice to guide preventive care [4,5]. Although ASCVD and CBI share overlapping vascular mechanisms, ASCVD scores were originally developed to predict overt macrovascular events rather than subclinical cerebral ischemic injury detectable on neuroimaging. Whether cumulative vascular risk captured by ASCVD estimation also reflects existing subclinical cerebrovascular injury remains unclear. In this community-based study, we investigated the association between 10-year ASCVD risk estimated by the Predicting Risk of Cardiovascular Disease EVENTs (PREVENT) equations and MRI-defined CBI in stroke-free middle-aged and older adults.
This cross-sectional analysis was conducted within the community-based PRECISE (PolyvasculaR Evaluation for Cognitive Impairment and vaScular Events) study (ClinicalTrials.gov NCT03178448), which enrolled 3,067 adults aged 50–75 years from urban and rural communities in China. Standardized interviews, physical examinations, and laboratory testing were performed at baseline [6]. The study protocol was approved by the ethics committees of Beijing Tiantan Hospital (KY2017-010-01) and Lishui Central Hospital (2016-42), and all participants provided written informed consent. The 10-year ASCVD risk was estimated using the PREVENT equations, including the base, urine albumin-creatinine ratio (UACR), and hemoglobin A1c (HbA1c) models. Participants were categorized into four risk strata (<5.0%, 5.0%–7.4%, 7.5%–19.9%, and ≥20.0%). The PREVENT-base equation includes routinely collected clinical predictors, including age, sex, total cholesterol, high-density lipoprotein cholesterol, systolic blood pressure, body mass index, estimated glomerular filtration rate, history of diabetes, smoking status, and use of antihypertensive or lipid-lowering medications. The PREVENT-UACR equation additionally includes UACR, and the PREVENT-HbA1c equation additionally includes HbA1c. Risk categories were defined according to commonly used 10-year ASCVD risk thresholds for clinical risk stratification. All participants underwent 3.0 T brain MRI using a standardized protocol, and CBI was centrally assessed by blinded neuroradiologists according to the Standards for Reporting Vascular Changes on Neuroimaging 2 criteria [7]. Multivariable logistic regression models, adjusted for clinically relevant covariates not directly incorporated in the PREVENT equations, were used to evaluate associations between PREVENT risk categories and CBI and its imaging phenotypes, with restricted cubic spline analyses applied to explore nonlinear relationships. Detailed MRI acquisition parameters for each sequence are provided in Supplementary Table 1 and imaging criteria for each imaging phenotype are provided in Supplementary Table 2, with their corresponding imaging examples shown in Supplementary Figure 1. All analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC, USA) and R version 4.6.0 (R Foundation for Statistical Computing, Vienna, Austria).
After excluding individuals with prior stroke or transient ischemic attack (TIA) and those with incomplete imaging data, 2,965 participants were included (Supplementary Figure 2). Among the included participants (mean age 61.05 years; 46.17% men), 284 (9.58%) had MRI-defined CBI. Participants with CBI were older and had a higher burden of vascular risk factors. Baseline characteristics by CBI status are summarized in Table 1. CBI prevalence increased from the low-risk through the borderline- and intermediate-risk categories across all three PREVENT equations. In the PREVENT-base model, CBI prevalence was 5.60%, 9.55%, and 16.67% in the low-, borderline-, and intermediate-risk groups, respectively. The high-risk groups were small (PREVENT-base, n=11; PREVENT-UACR, n=42; PREVENT-HbA1c, n=24), and estimates in these groups should therefore be interpreted cautiously (Supplementary Table 3). Similar gradients were observed in the PREVENT-UACR and PREVENT-HbA1c models (Supplementary Table 3).
In multivariable logistic regression models, higher PREVENT-ASCVD risk was shown to be associated with increased odds of CBI. Using the low-risk group as reference, the adjusted odds ratios (95% confidence intervals [CIs]) for the borderline- and intermediate-risk categories were 1.64 (1.16–2.32) and 2.97 (2.21–4.00) in the base model, 1.69 (1.19–2.39) and 2.65 (1.96–3.57) in the UACR model, and 1.69 (1.20–2.39) and 2.97 (2.20–4.01) in the HbA1c model, respectively. Associations for the borderline-and intermediate-risk categories remained statistically significant after adjustment for current drinking status, homocysteine levels, chronic lung disease, tumor history, and educational attainment across all three PREVENT models. In the high-risk category, the association remained significant in the PREVENT-UACR model, whereas estimates were not statistically significant in the PREVENT-base and PREVENT-HbA1c models (Figure 1 and Supplementary Table 3). In sensitivity analyses using an alternative PREVENT risk categorization, the association between higher PREVENT-estimated ASCVD risk and MRI-defined CBI remained materially unchanged (Supplementary Table 4). Restricted cubic spline analyses demonstrated a nonlinear association between continuous PREVENT-ASCVD risk and CBI prevalence (Figure 2). The increase in CBI odds was particularly steep when the estimated 10-year ASCVD risk was below approximately 10%, after which the curve attenuated. This suggests that individuals traditionally considered at “low-to-moderate” cardiovascular risk may already carry a substantial burden of covert cerebrovascular injury. Similar graded associations were observed across CBI imaging phenotypes. As shown in Supplementary Table 5, participants in the intermediate-risk group had more than twofold higher odds of chronic infarctions and over threefold higher odds of cavitatory lesions. Although acute infarctions were infrequent, a consistent upward trend was observed across risk categories. These findings demonstrate that global cardiovascular risk estimated by the PREVENT equations is closely associated with covert cerebrovascular injury detectable by MRI. The consistency of associations across imaging phenotypes further supports the link between systemic vascular risk and subclinical cerebral ischemia.
Adjusted OR with 95% CI for presence of CBI according to PREVENT-ASCVD risk score categories. OR for presence of CBI according to PREVENT-ASCVD risk scores categories adjusted for education, drinking, level of homocysteine, history of chronic lung disease and cancer. OR and 95% CI (vertical bars) are shown for the borderline, intermediate, and high-risk groups, with the low-risk group as reference. The scores from all three models of PREVENT-ASCVD (base, UACR and HbA1c) were categorized as low (<5.0%), borderline (5.0%–7.4%), intermediate (7.5%–19.9%) or high (≥20.0%) risk. (A) PREVENT-base model. (B) PREVENT-UACR model. (C) PREVENT-HbA1c model. OR, odds ratio; CI, confidence interval; CBI, covert brain infarction; PREVENT, the Predicting Risk of Cardiovascular Disease EVENTs; ASCVD, atherosclerotic cardiovascular disease; UACR, urine albumin-creatinine ratio; HbA1c, hemoglobin A1c.
Association between PREVENT-ASCVD risk scores and overall CBI risk. (A) PREVENT-base model, incorporating traditional cardiovascular risk factors. (B) PREVENT-UACR model, which additionally includes UACR. (C) PREVENT-HbA1c model, which additionally includes glycated hemoglobin (HbA1c). The diagram presents ORs as solid lines and 95% CIs as shaded areas obtained from restricted cubic spline regressions. The histograms show the density distribution of PREVENT risk values and correspond to the right y-axis. The dashed line at OR 1.0 serves as the reference. Data were fitted using logistic regression models with restricted cubic spline with 3 knots (the 10th, 50th, and 90th percentiles) for PREVENT-ASCVD risk, adjusting for education, drinking, level of homocysteine, history of chronic lung disease and cancer. P for overall and P for nonlinear were derived from this model. OR, odds ratio; CI, confidence interval; PREVENT, the Predicting Risk of Cardiovascular Disease EVENTs; UACR, urine albumin-creatinine ratio; HbA1c, hemoglobin A1c; ASCVD, atherosclerotic cardiovascular disease; CBI, covert brain infarction.
Many established determinants of CBI, including advanced age, hypertension, diabetes, dyslipidemia, and elevated homocysteine, overlap with major components of ASCVD risk, suggesting shared pathways of chronic vascular injury [8,9]. Previous studies have linked traditional cardiovascular risk scores to MRI markers of cerebral small vessel disease [10,11]. Previous work from the PRECISE cohort showed that conventional 10-year ASCVD risk scores, including the FRS, PCE, China-PAR, and SCORE2, were associated with CSVD burden and individual MRI markers [12]. The present study extends this work by focusing on the more recently developed PREVENT equations and MRI-defined CBI and its imaging phenotypes in participants without a prior history of stroke or TIA, which provides complementary evidence linking contemporary cardiovascular risk estimation with MRI-defined covert brain infarction.
From a clinical perspective, these findings have important implications. Because CBI is typically identified only through neuroimaging yet confers substantial risk for future stroke and cognitive decline, PREVENT-based risk assessment may provide a practical framework for understanding the overlap between systemic cardiovascular risk and covert cerebrovascular injury. Among the three models, the PREVENT base equation may be the most readily applicable in routine practice because it relies on routinely collected variables. The comparable associations across the PREVENT-base, PREVENT-UACR, and PREVENT-HbA1c equations may reflect that the base equation already incorporates major CBI-related vascular risk factors, including age, blood pressure, lipid measurements, body mass index, kidney function, diabetes, smoking status, and relevant medication use. Thus, UACR and HbA1c may have provided limited additional information for CBI risk stratification in this community-based sample, possibly because of overlap with cardiometabolic variables already included in the base equation. Future longitudinal and comparative studies are needed to determine whether PREVENT-based cardiovascular risk assessment provides incremental clinical or predictive value over conventional cardiovascular risk scores for CBI. This study is limited by its cross-sectional design, single-region population, and absence of direct predictive-performance comparisons between PREVENT and traditional risk scores; therefore, it cannot establish the incremental clinical or predictive value of PREVENT for identifying CBI. Nevertheless, the integration of a contemporary ASCVD risk model with standardized MRI assessment in a community-based cohort provides novel evidence linking global cardiovascular risk and covert cerebral ischemic injury.
In summary, higher PREVENT-estimated ASCVD risk was associated with an increased likelihood of CBI in asymptomatic adults, supporting a close association between global cardiovascular risk and covert cerebrovascular injury.
Supplementary materials
Supplementary materials related to this article can be found online at https://doi.org/10.5853/jos.2026.00766.
MRI parameters of the PRECISE study
Imaging criteria for imaging phenotypes of covert brain infarction
Associations between PREVENT-ASCVD risk scores categories and the presence of CBI
Associations between revised PREVENT-ASCVD risk scores categories and the presence of CBI
Associations between PREVENT-ASCVD risk scores categories and the presence of CBI imaging phenotypes
Imaging phenotypes of covert brain infarction. The arrows indicate representative lesions of the corresponding covert brain infarction imaging phenotypes. (A) Cavitatory infarction on FLAIR. (B) Acute cerebral infarction on DWI. (C) Chronic infarction: combined gray and white matter lesions on T2-weighted and FLAIR images. (D) Chronic infarction: large subcortical non-cavitated infarcts on FLAIR images. FLAIR, fluid-attenuated inversion recovery; DWI, diffusion weighted imaging.
Flowchart of participant selection. PRECISE, PolyvasculaR Evaluation for Cognitive Impairment and vaScular Events; CBI, covert brain infarction; TIA, transient ischemic attack.
Notes
Funding statement
This work was supported by grants from Beijing High-level Public Health Technical Personnel Construction Project (Discipline leader-03-12), Noncommunicable Chronic Diseases-National Science and Technology Major Project (2023ZD0504800, 2023ZD0504801, 2023ZD0504802, 2023ZD0504803, 2023ZD0504804), and grants from AstraZeneca Investment (China) Co., Ltd.
Conflicts of interest
The authors have no financial conflicts of interest.
Author contribution
Conceptualization: Peiyuan Gao, Yilong Wang, Yuesong Pan. Study design: Peiyuan Gao, Yilong Wang, Yuesong Pan. Methodology: Peiyuan Gao, Yilong Wang, Yuesong Pan. Data acquisition: Xuan Wang, Lijie Ren, Mengyuan Zhou, Ruinan Zhang, Xueli Cai, Yingying Yang, Jingping Sun, Yuesong Pan. Statistical analysis: Peiyuan Gao. Supervision: Yilong Wang, Yuesong Pan. Writing—original draft: Peiyuan Gao. Writing—review & editing: all authors. Approval of final manuscript: all authors.
Acknowledgments
The authors thank the staff and participants of the PRECISE study for their important contributions.
Deidentified participant data and statistical analysis codes are available from the PRECISE study coordinating center upon reasonable request to the corresponding author.
