Both chronological age and individual differences in aging are the two indispensable components for predicting...

Due to the lack of a gold standard for biological age (BA), existing studies usually adopt chronological age (CA) as the training label when constructing BA prediction models using supervised artificial intelligence algorithms. However, CA cannot reflect individual differences in aging, leading to the frequently observed age delta correlation (ADC) phenomenon. This work proposes a new method for predicting vascular biological age using arteriosclerosis detection data. A novel index, relative individual risk difference (RIRD), is introduced to quantify individual differences in aging. Based on RIRD, an approximate absolute health group is selected from the healthy population. The health group is used to establish a supervised regression model for vascular biological age prediction, while the approximate absolute health group is further employed to obtain a corrected vascular biological age. RIRD is defined as the difference between an individual's vascular biological age and the mean vascular biological age of the health group. Compared with the widely used Klemera-Doubal method (KDM), the corrected vascular biological age shows comparable ability in detecting non-healthy individuals, while outperforming the uncorrected vascular biological age. Moreover, vascular biological age without considering RIRD exhibits a significant ADC phenomenon, whereas both the corrected vascular biological age and KDM effectively eliminate this bias. These results indicate that CA and individual differences in aging are two indispensable components of BA. Therefore, incorporating individual aging heterogeneity is essential when constructing BA prediction models using supervised artificial intelligence algorithms.