Quantifying cardiovascular autonomic aging with machine learning.

Area of Science:

• Cardiovascular physiology
• Biomarkers
• Machine learning in medicine

Background:

• Precision medicine and aging research increasingly utilize machine learning (ML).
• Autonomic cardiovascular function provides insights into physiological aging.
• Existing metrics may not fully capture age-related cardiovascular decline.

Purpose of the Study:

• Introduce the cardiovascular autonomic age (CAA) gap, a novel ML-based metric.
• Quantify the deviation between ML-estimated biological age and chronological age.
• Evaluate the CAA gap's association with cardiovascular risk.

Main Methods:

• Derived 29 autonomic indices from high-resolution ECG and continuous BP recordings in 1,060 healthy individuals.
• Trained a Gaussian process regression model on 879 participants to estimate biological age (CAA).
• Calculated the CAA gap as the difference between CAA and chronological age, validated in risk-stratified cohorts.

Main Results:

• The high cardiovascular risk (CVR) group exhibited a significantly increased CAA gap (+11 yr) compared to the low-CVR group (-1 yr).
• CAA correlated positively with Framingham risk score (r=0.42, P<0.001).
• Elevated CAA was consistently observed in the high-CVR group across risk thresholds.

Conclusions:

• The CAA gap serves as a sensitive and interpretable indicator of cardiovascular risk and accelerated physiological aging.
• This novel metric holds potential for early detection and longitudinal assessment of cardiovascular health.
• ML-based autonomic markers offer new avenues for understanding aging processes.