Evaluating multimodal autonomic correlations in pediatric epilepsy using wearable biosignals

OBJECTIVE: Antiseizure medication (ASM) may affect autonomic nervous system (ANS) activity in patients with epilepsy (PWE). We examined the relationship between ASM dosage and multimodal correlations among ANS signals recorded from wearables in pediatric PWE. METHODS: We evaluated evening (ASM intake from 5 p.m. to 3 a.m.) multimodal recordings (heart rate [HR], electrodermal activity [EDA], temperature [TEMP], and respiratory rate [RR]) from wearables (Empatica E4) worn by pediatric PWE during long-term monitoring at Boston Children's Hospital between 2015 and 2021. Within-patient comparisons were performed in two groups: patients with both high- and no-dose ASM days, and with both high- and low-dose ASM days. Multimodal interactions were assessed using principal component and canonical correlation analysis, and repeated measure analyses of variance with time and dose as factors. RESULTS: Of the 52 patients (median age = 12.8 years), 34 total patients had both high- and low-dose ASM days, and 24 total patients had both high- and no-dose days. An interaction between dose and time emerged in the high- versus no-dose comparison (p = .002), indicating divergent trajectories of multimodal autonomic correlations across medication states; correlations increased on high-dose days and decreased on no-dose days. EDA increased (p = .003) and HR decreased (p = .036) from baseline to peak window for patients with both high- and low-dose days. No time effects or dose-time interactions were found for TEMP and RR. Subanalyses by ASM mechanism of action showed no differential effects on individual ANS measures. Cox modeling showed a dose effect on time to seizure (chi-squared = 6.98, p = .031), with higher hazard on low- versus high-dose days (p < .01) and no difference between no- and high-dose days (p = .626). SIGNIFICANCE: ASM dosage was related to multimodal autonomic correlations, suggesting central autonomic regulation and seizure vulnerability. Wearable-based monitoring of these correlations could support seizure risk assessment and inform personalized treatment strategies.