Abstract
PURPOSE: Paroxysmal sympathetic hyperactivity (PSH) is a severe complication of acquired brain injuries (ABIs), characterized by sudden autonomic surges that exacerbate clinical outcomes. Its pathophysiology remains debated, and early biomarkers are lacking. This study aims to investigate autonomic changes preceding PSH and assess the feasibility of predictive modeling using heart rate variability (HRV). METHODS: Continuous electrocardiogram (ECG) recordings were obtained from six male patients with disorders of consciousness (DoC), including unresponsive wakefulness syndrome and minimally conscious state. A total of 24 PSH episodes and 24 matched control (noPSH) events were analyzed. HRV metrics, including entropy measures and power spectral density (PSD), were evaluated. A support vector machine (SVM) classifier was implemented to differentiate PSH from control events and to predict PSH onset. RESULTS: PSH events were associated with significant heart rate increases, reduced entropy-based complexity, and decreased PSD in both low-frequency (LF) and high-frequency (HF) bands. An increased very-low-frequency (VLF)/(LF + HF) ratio suggested potential involvement of the renin-angiotensin-aldosterone system (RAAS) in PSH pathogenesis. The SVM classifier achieved perfect classification during the event. In addition, 10 min prior to onset, the model reached 67% sensitivity, 100% specificity, and 83% balanced accuracy. CONCLUSIONS: HRV analysis reveals distinct autonomic signatures preceding PSH and suggests, as a working hypothesis, that dysregulation of the RAAS may play a role. However, VLF power is influenced by multiple mechanisms and cannot be considered a specific or exclusive marker of RAAS activity. SVM-based predictive modeling offers a promising tool for PSH detection, providing a basis for investigating autonomic/neuroendocrine regulation, including RAAS.