A Predictive Model of Cardiovascular Aging by Clinical and Immunological Markers Using Machine Learning

Background/Objectives: Aging and immune mechanisms play a key role in the development of cardiovascular disease (CVD), especially in the context of chronic inflammation. Therefore, in order to detect early aging in the elderly, we have developed a prognostic model based on clinical and immunological markers using machine learning. Methods: This paper analyzes the relationships between immunological markers, clinical parameters, and lifestyle factors in individuals over 60 years of age. A machine learning (ML) model including random forest, logistic regression, k-nearest neighbors, and XGBoost was developed to predict the aging rate and risk of CVD. Correlation anal is revealed significant associations between immune markers (CD14+, HLA-DR, IL-10, CD8+), clinical parameters (BMI, coronary heart disease, hypertension, diabetes), and behavioral factors (physical activity, smoking, alcohol). Results: The results of the study confirm that systemic inflammation, as reflected by markers such as CD14+, HLA-DR, and IL-10, plays a central role in the pathogenesis of aging and related diseases. CD14+ shows a moderate positive correlation with post-infarction cardiosclerosis, accounting for 37%. HLA-DR correlates with body mass index at 39%. A negative association between IL-10 level and BMI was also found, where the correlation reaches 52% (r = -0.52). The level of CD8+ cells shows a negative correlation with smoking and their number, being 40%. Training was performed on clinical and immunological data and models were evaluated using accuracy, ROC-AUC, and F1-score metrics. Among all the trained models, the XGBoost model performed best, achieving an accuracy of 91% and an area under the ROC curve (AUC) of 0.8333. Conclusions: The study reveals significant correlations between immunological markers and clinical parameters, which allows the assessment of individual risks of premature cardiovascular aging. R (version 4.3.0) and specialized libraries for correlation matrix construction and visualization were used for data analysis, and Python (version 3.11.11) was used for model development and training.