The autonomic age gap: a machine learning approach to assess biological-calendar age deviations
Preprint
Abstract
Machine learning has emerged as a valuable tool in precision medicine and aging research. Here, we introduce the autonomic age gap , a novel metric quantifying the individual deviation between machine-learning–estimated biological age and chronological age, based on autonomic nervous system function. We collected high-resolution electrocardiograms and continuous blood pressure recordings at rest from 1,012 healthy individuals. From these signals, 29 autonomic indices were extracted, encompassing time-, frequency-, and symbol-domain heart rate variability, cardiovascular coupling, pulse wave dynamics, and QT interval features. Based on those parameters, a Gaussian process regression model was trained on 879 participants to estimate chronological age referred to as autonomic age . The model was used to estimate the deviation from expected healthy aging, the autonomic age gap , in an independent validation set and two test sets stratified by cardiovascular risk (CVR) using the Framingham. The was evaluated via the autonomic age gap. High CVR individuals had a significantly increased autonomic age gap of 9.7 years compared to the low CVR group and the validation set. In contrast, the low CVR group had a negative age gap of -2.2 years on average. Predictions in the validation sample closely matched calendar age with a deviation below 0.5 years. Additionally, in the high-risk group, the slope of predicted versus actual age suggested accelerated physiological aging. These findings highlight the autonomic age gap as a sensitive and interpretable marker of cardiovascular risk and aging, offering potential clinical utility for early risk detection and longitudinal health monitoring.Keywords
This publication has 48 references indexed in Scilit:
- Decoding aging clocks: New insights from metabolomicsCell Metabolism, 2024
- Hallmarks of aging: An expanding universeCell, 2023
- Assessing the rate of aging to monitor aging itselfAgeing Research Reviews, 2021
- Commentary: Correction procedures in brain-age predictionNeuroImage: Clinical, 2020
- Artificial Intelligence Is Becoming NaturalCell, 2018
- Extended algorithm for real-time pulse waveform segmentation and artifact detection in photoplethysmogramsSomnologie, 2017
- Blood Pressure Variability: A Novel and Important Risk FactorCanadian Journal of Cardiology, 2013
- Using Body Mass Index Data in the Electronic Health Record to Calculate Cardiovascular RiskAmerican Journal of Preventive Medicine, 2012
- Arterial Age as a Function of Coronary Artery Calcium (from the Multi-Ethnic Study of Atherosclerosis [MESA])The American Journal of Cardiology, 2009
- QT variabilityJournal of Electrocardiology, 2003