An FDA-cleared smartwatch app detected nearly every tonic-clonic seizure in a Phase III trial while producing just one false alarm every 12 days, potentially addressing one of the biggest challenges in wearable epilepsy monitoring.
Study: Phase III Trial of EpiWatch for Tonic-Clonic Seizure Detection in Children and Adults. Image credit: VI-stock studio/Shutterstock.com
A Phase III clinical trial evaluating the effectiveness of the smartwatch-based app EpiWatch concluded that it could detect tonic-clonic seizures (TCS) relatively early and had a low false alarm rate (FAR). However, its real-world performance remains to be tested. The study was published in the journal Neurology Open Access.
Why tonic-clonic seizures raise SUDEP risk
Tonic-clonic seizures (TCS), which include both generalized and focal-to-bilateral seizures, are among the most dangerous forms of epilepsy. Beyond the immediate risks of injury and prolonged seizures, they are strongly associated with sudden unexpected death in epilepsy (SUDEP), particularly in people who live or sleep alone. After a seizure, patients often experience a period of immobility, muscle flaccidity, and breathing dysfunction, including, in some cases, terminal apnea, which are thought to contribute to this elevated risk.
EpiWatch brings seizure detection to consumer wearables
The threefold lower risk of SUDEP among TCS patients who share a bedroom with someone else suggests the benefit of early detection and timely intervention. Accordingly, the International League Against Epilepsy (ILAE) recommends using wearable devices to detect TCS and generate alerts.
However, given the stigma associated with epilepsy, devices intended to be worn only by epilepsy patients are less popular. Moreover, high FARs (often outnumbering true alarms) may induce caregiver fatigue. Given this need, researchers at Johns Hopkins developed the EpiWatch app, which uses a seizure-detection algorithm on the widely used Apple Watch to identify TCS in children and adults with epilepsy.
The app was licensed to EpiWatch, Inc., which funded the clinical trial. Notably, the authors note that this is the first Apple Watch seizure-monitoring app to undergo a Phase III clinical trial.
EpiWatch, Inc., did not participate in the study, and Johns Hopkins researchers were involved under appropriate precautions to mitigate potential conflicts of interest. The paper notes that EpiWatch was developed at Johns Hopkins, two study authors are shareholders in the company, and the trial incorporated blinding and conflict-management procedures to reduce potential bias. Based on this trial, the app received Food and Drug Administration (FDA) 510(k) clearance for marketing as the EpiWatch – Seizure Monitor, a prescription app available on the Apple App Store.
Phase III trial tests smartwatch seizure detection
The current trial was a prospective, multicenter Phase III study designed to assess the diagnostic accuracy of the app. It involved six epilepsy monitoring units (EMUs) over a two-year period. The trial included 242 participants, either children aged 5 years or older or adults aged 22 years or older (mean age 22.7 years), with a history of TCS or clinically likely to have TCS. All participants were being monitored by video and electroencephalography (EEG).
The aim was to evaluate the reliability of TCS detection by the wearable, using movement and physiological signals. The app can generate caregiver alerts upon seizure detection, although alerts were disabled during the trial to maintain blinding. The app detections were compared with clinically confirmed TCS events by an independent expert panel.
EpiWatch detects nearly every tonic-clonic seizure
Approximately 34% of the participants had seizure events with motor activity during the monitoring period. Thirty-seven participants had a TCS, with 27 of them experiencing only one seizure. Of the TCS, 19 occurred during sleep.
The researchers found that EpiWatch detected 46 of 47 TCS events, thus showing a high sensitivity of 98%. For the one seizure that was missed by EpiWatch, the caregiver had kept the participant’s arm on which the watch was worn under restraint. The adjusted positive percent agreement (PPA) was 94%, indicating a high degree of agreement between the detected TCS and clinically confirmed TCS used as the reference.
False alarms occur only once every 12 days
Of the total 16,189 hours of monitoring, EpiWatch produced 56 false alarms, equivalent to one every approximately 12 days. According to the authors, this is approximately one-tenth the false alarm rate reported for other published prospective TCS-monitoring devices.
About 38% of the false alarms were motor seizures that did not meet TCS criteria, and the others originated from movements during video games or repetitive movements. Excluding detections associated with these non-TCS motor seizures, the FAR for non-seizure-related detections is one false alarm every 18.7 days.
The authors estimated that in a population experiencing a median of three TCSs per month, the positive predictive value (PPV) would be approximately 0.55 for TCS and still higher for any seizure. This means that true seizures would be expected to outnumber false alarms. This reduces the risk of caregiver fatigue, where frequent detections are ignored. The FAR remained consistently low across age groups and did not vary by sex or race, unlike other wearables, which perform better in adults.
Seizures detected within half a minute
The median latency of TCS detection (time to detect a TCS) was 31.5 seconds after clinical onset, allowing for relatively rapid caregiver alerting. This is comparable to that of other wearables, well within the median duration of a TCS. Just as important, it falls well within the recommendation for seizure detection latency of within one minute.
Every nighttime tonic-clonic seizure was detected
All TCS occurring during sleep were successfully detected. All false alarms during sleep were associated with seizure activity. This indicates that the system may work particularly well for reliable seizure detection during nighttime sleep, especially since no normal sleep behavior triggered a false alarm during the study.
Other advantages
The integration of EpiWatch into the Apple Watch is likely to appeal to epilepsy patients because it is not a disease-specific device and requires no additional training. It is compatible with older or refurbished Apple Watch models, according to the authors, potentially lowering cost barriers to adoption.
Unrelated advantages include the Apple Watch's ability to host software that could map seizure frequency and patterns, medication logs, seizure triggers, and other illnesses linked to seizures. However, caregivers must have Wi-Fi or cellular phone access to receive alerts. Like any wearable, EpiWatch's effectiveness depends on the patient keeping it charged and wearing it.
Real-world performance still needs validation
A key limitation of this trial is the controlled EMU environment, which may differ significantly from actual use conditions. Actual data on SUDEP risk after EpiWatch usage would require large, diverse real-world studies with long follow-up periods. Moreover, the study addressed only TCS in individuals aged 5 years and older, limiting its generalizability.
Smartwatch monitoring could strengthen epilepsy care
The researchers emphasize that wearables and apps are not replacements for clinical evaluation or standard epilepsy management. However, they could be a valuable adjunct to epilepsy care, especially for high-risk patients requiring continuous supervision, supporting earlier intervention during seizures. Long-term real-world studies are needed to determine how performance translates outside the EMU setting.
Download your PDF copy by clicking here.
Unhealthy snacks drive disruptive behavior in young children