A data-driven multiple linear regression model for cardiovascular homeostatic regulation in health, ageing, and disease

Heart rate variability (HRV) is a widely used noninvasive marker of cardiovascular and general health risk, whereas blood pressure variability (BPV) remains comparatively underexplored. Under adverse conditions such as aging or disease, HRV typically decreases while BPV increases. However, most studies assess these metrics independently, neglecting their potential multivariate interactions. In this work, we analyze two datasets: the autonomic aging (AA) dataset (n= 1,121 healthy individuals stratified into age groups) and the type-2 diabetes (T2D) dataset (n = 75, including controls and short- and long-term T2D patients). We examine variations in mean heart rate (HR), HRV, mean blood pressure (BP), and BPV across health states. Univariate analyses showed that HRV (sdIBI) was the most sensitive metric, consistently decreasing with aging and disease, whereas BPV (sdSBP) differed mainly in older individuals or long-term T2D. Bivariate and multivariate models were constructed using correlation matrices and linear regression, revealing two distinct correlation patterns: one characteristic of youth and health, dominated by interactions between HRV and mean HR, and another associated with aging and disease, characterized by absent correlations or a dependence of BPV on HRV. These findings align with principles of homeostatic regulation, where BP acts as a regulated variable and HR as an effector. Overall, the results indicate that HRV and BPV cannot be fully understood in isolation; their joint analysis reveals shared regulatory dynamics across health states and supports the use of multivariate physiological models in clinical practice.