Body movements as biomarkers: Machine Learning-based prediction of HPA axis reactivity to stress

Body movements and posture provide valuable insights into stress responses, yet their relationship with endocrine biomarkers of the stress response remains underexplored. This study investigates whether movement patterns during the Trier Social Stress Test (TSST) and the friendly-TSST (f-TSST) can predict cortisol reactivity. Using motion capturing, movement data from 41 participants were analyzed alongside salivary cortisol responses. Machine learning models achieved a classification accuracy of 65.2 % for distinguishing cortisol responders from non-responders and a regression mean absolute error of 2.94 nmol/l for predicting cortisol increase. Findings suggest that movement dynamics can serve as proxies of endocrine stress responses, contributing to objective, non-invasive stress assessment methods.