AI-Induced Vascular Ages Are a Measurable Residual Risk for Cardiovascular Diseases in the Japanese Population

Abstract

Background: Cardiovascular diseases (CVDs) remain a leading cause of morbidity and mortality, despite advances in treatment. Early detection of vascular aging is critical, as preclinical atherosclerosis often remains undiagnosed. AI-determined vascular age, originally developed using carotid-femoral pulse wave velocity (cf-PWV), may help to identify individuals at elevated risk. This study aimed to evaluate the clinical utility of an alternative AI-determined vascular age model based on the arterial velocity pulse index (AVI) and arterial pressure volume index (API) in a Japanese hospital-based cohort. Methods: This retrospective, exploratory study analyzed electronic health records of 408 patients from Yokohama City University Hospital. This study was approved by the Clinical Research Ethics Committee (approval numbers: B180300040, F240500007), and patient consent was obtained through an opt-out process. AI-determined vascular age was estimated using a Generalized Additive Model (GAM) with backward stepwise regression, substituting cf-PWV with AVI and API. Correlations with chronological age were assessed, and comparisons of cardiovascular and renal function markers were performed across age-stratified groups. Results: AI-determined vascular age showed a strong correlation with chronological age (p < 0.05). Significant differences were observed in cardiac diastolic function parameters, B-type natriuretic peptide (BNP), and estimated glomerular filtration rate (eGFR) between the highest and lowest quintiles of AI-determined vascular age. Conclusions: AI-determined vascular age using AVI and API appears to be a feasible surrogate for cf-PWV in clinical settings. This index may aid in stratifying vascular aging and identifying individuals who could benefit from early cardiovascular risk management.

Keywords: arterial stiffness; cardiovascular diseases; computer-aided vascular age; vascular age.

Figure 1

Flowchart showing inclusion and exclusion processes of the current study for Cohort I.

Figure 2

Results of stepwise regression analysis with chronological age as the dependent variable (a) and the calculation results of the Variance Inflation Factor (VIF) to assess multicollinearity (b).

Figure 3

Results of the selected explanatory variables transformed using spline smoothing. Horizontal lines show each transformed variables and vertical lines show chronological age. The shaded area represents the 95% confidence interval. Whether the value of the smoothing function is positive or negative indicates the direction of the effect of the variable on the dependent variable (chronological age).

Figure 4

Flowchart showing inclusion and exclusion processes of the current study for Cohort II.

Figure 5

These figures show the results of AI vascular age in Cohort I. The average and standard error of chronological age is 69.0 ± 12.2, and that of AI vascular age is 69.0 ± 8.7, respectively. (a) In 52.7% of the cases, chronological age was higher than AI vascular age, whereas in 42.3% of the cases, AI vascular age was higher than chronological age (b). (c) Linear regression relationships between chronological age and AI vascular age. (d) Finally, there is a significant relationship between ΔAge (chronological age − AI vascular age) and chronological age.

Figure 6

Comparison of cardiovascular markers and eGFR between the oldest 20% (top) and the youngest 20% (bottom) for chronological age and AI vascular age in Cohort I using the U-test.

Figure 7

Results of AI vascular age in Cohort II. The average and standard error of chronological age is 66.9 ± 12.8, and that of AI vascular age is 70.4 ± 9.1. (a) In 52.7% of the cases, chronological age was higher than AI vascular age, whereas in 42.3% of the cases, AI vascular age was higher than chronological age (b). (c) Linear regression relationships between chronological age and AI vascular age. Finally, there is a significant relationship between ΔAge (chronological age − AI vascular age) and chronological age (d).

Figure 8

Comparison of BNP and eGFR between the oldest 20% (top) and the youngest 20% (bottom) for chronological age and AI vascular age in Cohort II using the U-test.