Epiwatch for tonic-clonic seizure detection in children and adults

Krauss GL, Elizebath R, Shah S, et al. Phase III trial of epiwatch for tonic-clonic seizure detection in children and adults https://www.neurology.org/doi/10.1212/WN9.0000000000000111. Neurol Open Access. 2026; Epub 2026 May 27.

Abstract

Background and Objectives

Risks of sudden unexpected death in epilepsy are high in individuals with uncontrolled tonic-clonic seizures (TCSs), particularly in those who sleep alone. Wearable seizure detection devices can alert caregivers to provide timely intervention; however, current devices risk stigma and have high false alarm rates (FARs) that discourage consistent use and caregiver responses. EpiWatch is a seizure detection application (app) developed for the Apple Watch, designed to minimize FAR without compromising detection sensitivity or latency. We evaluated the sensitivity and FAR of EpiWatch for detecting TCSs in children and adults undergoing video-EEG monitoring.

Methods

We conducted a prospective, multicenter Phase III diagnostic accuracy study at 6 epilepsy monitoring units (EMUs) from September 2021 to October 2023. Children and adults aged 5 years and older with a history of TCSs or clinical potential for TCSs and who underwent video-EEG monitoring were eligible. Participants wore the EpiWatch device on the wrist contralateral to their seizure focus. EpiWatch detections were compared with TCS events classified by an independent Central Reader Panel of epileptologists blinded to device output. Co-primary end points were sensitivity and FAR per 24 hours. Secondary end points included detection latency and performance during sleep. Enrollment was slowed during COVID-19 restrictions, which extended the recruitment period.

Results

A total of 242 participants were enrolled (mean age 22.7 years; 54.1% female; 60.7% aged 5–21). EpiWatch detected 46 of 47 panel-verified TCSs; 1 seizure was missed when a caregiver restrained the participant's arm. The overall sensitivity was 98% (95% CI 95%–100%). There were 56 false alarms during 16,189 hours of monitoring, with a FAR of 0.08 per 24 hours (95% CI 0.02–0.12), equivalent to 1 false alarm every 12.4 days. All age groups had similarly low FARs. The median detection latency was 31.5 seconds. During sleep, all TCSs were detected, and all false alarms from sleep were associated with seizure activity. No adverse events occurred.

Discussion

EpiWatch detected TCSs with high sensitivity and a FAR approximately one-tenth that of other published devices. These findings were obtained in a controlled EMU environment; real-world performance may differ.

Classification of Evidence
This study provides Class I evidence that, in patients with a history of TCSs, the EpiWatch accurately detects TCSs, with a low FAR. EPW001.