Actigraphic Correlates of Neuropsychiatric Symptoms in Adults with Focal Epilepsy.

Abstract

Disability in patients with epilepsy (PWE) is multifactorial: beyond seizure frequency/severity, PWE are prone to a range of neuropsychiatric, cognitive, and somatic comorbidities that significantly impact quality of life. Here, we explored how variations in seizure severity and the burden of self-reported somatic/neuropsychiatric symptoms correlate with disruptions to 24h activity patterns (rest-activity rhythms, RARs), determined through wrist accelerometry/actigraphy. Multiday wrist-actigraphy recordings were obtained from 59 adult patients with focal epilepsy (44% male, ages 18-72), who contemporaneously responded to validated psychometric instruments to measure anxiety, depression, sleepiness, and somatic symptoms. We conducted a similar in silico psychometric-actigraphic correlation in a publicly available dataset of 1747 Hispanic subjects (35% male, ages 18-65) from the Study of Latinos (SOL) Sueño Ancillary Study. RARs were analyzed via a sigmoidally-transformed cosine model (quantifying amplitude, steepness, acrophase and robustness) and non-parametric measures to estimate RAR stability, fragmentation, and sleep. Compared with matched SOL subjects, RARs from PWE subjects featured a significantly lower amplitude, a wider rest phase and significantly more total daily sleep. Within PWE, similar RAR distortions were associated with seizure intractability and/or anticonvulsant polytherapy, whereas high anxiety, depression, and somatic symptom scores were associated with lower RAR robustness and acrophase delay. We applied the SOL dataset to train logistic regression models to dichotomously classify subjective anxiety, depression and sleepiness symptoms using demographic and RAR parameters. When tested on PWE, these models predicted prevalent anxiety and depression symptom burden (accuracy ~70%) but failed to predict subjective sleepiness. Together, these results demonstrate that RAR features may encode prevalent depression and anxiety symptoms in patients with focal epilepsy, potentially offering wearable-derived endpoints to adjunct clinical care and drug/device trials. With larger PWE-specific actigraphic-psychometric datasets, we may identify RAR signatures that may more precisely correlate with varying seizure frequency, the burden of anticonvulsant therapy and prevalent mood/anxiety symptoms.