AHAㅣ 2024-11-11
Abstract
Introduction: Cardiovascular disease (CVD) remains the leading cause of mortality globally, necessitating improved risk stratification and preventive strategies. Reti-CVD, a deep learning model, predicts CVD risk by analyzing retinal images to detect subtle vascular changes indicative of coronary artery calcium presence. This study evaluates the efficacy of statin therapy in individuals identified as high-risk by Reti-CVD using propensity score matching and survival analysis.
Hypothesis: Individuals with high Reti-CVD scores will benefit significantly from statin therapy, resulting in a reduced incidence of cardiovascular events compared to those not receiving statins.
Methods: Data from the UK Biobank were utilized to perform 1:1 propensity score matching, comparing statin users and non-users. Matching was based on age, gender, race, BMI, systolic blood pressure, diastolic blood pressure, medication use for blood pressure, diabetes status, smoking status, alcohol consumption, and socioeconomic status. Within a cohort identified as high-risk by a deep learning model, survival analysis of a mean follow-up period of 9.83 years for CVD outcomes was conducted, with the hazard ratio calculated for statin users versus non-users.
Results: In the unmatched cohort, significant differences were observed in several variables including age and gender between statin users (n=3,008) and non-users (n=42,373) as shown in Table 1. Post-matching, the cohort comprised 3,008 statin users and 3,008 non-users, with standardized mean differences significantly reduced for most variables. For the high Reti-CVD cohort, the hazard ratio for CVD outcomes in the statin group was 0.76 (95% CI: 0.50-1.14), indicating a 24% reduction in the risk of cardiovascular events compared to non-users. In comparison, the low to medium Reti-CVD cohort had a hazard ratio of 1.22 (95% CI: 0.90-1.68).
Discussion: Findings suggest that individuals with high Reti-CVD scores likely benefit from statin therapy, as indicated by a trend towards reduced cardiovascular events. Reti-CVD shows potential in refining risk stratification for guiding statin therapy decisions. The integration of deep learning models like Reti-CVD in clinical practice could provide targeted interventions for high-risk populations.
