Detection of generalized tonic–clonic seizures by a wireless wrist accelerometer: A prospective, multicenter study

Our objective was to assess the clinical reliability of a wrist-worn, wireless accelerometer sensor for detecting generalized tonic-clonic seizures (GTCS). Seventy-three consecutive patients (age 6-68 years; median 37 years) at risk of having GTCS and who were admitted to the long-term video-electroencephalography (EEG) monitoring unit (LTM) were recruited in three centers. The reference standard was considered the seizure time points identified by experienced clinical neurophysiologists, based on the video-EEG recordings and blinded to the accelerometer sensor data. Seizure time points detected real-time by the sensor were compared with the reference standard. Patients were monitored for 17-171 h (mean 66.8; total 4,878). Thirty-nine GTCS were recorded in 20 patients. The device detected 35 seizures (89.7%). In 16 patients all seizures were detected. In three patients more than two thirds of the seizures were detected. The mean of the sensitivity calculated for each patient was 91%. The mean detection latency measured from the start of the focal seizure preceding the secondarily GTCS was 55 s (95% confidence interval [CI] 38-73 s). The rate of false alarms was 0.2/day. Our results suggest that the wireless wrist accelerometer sensor detects GTCS with high sensitivity and specificity. Patients with GTCS have an increased risk for injuries related to seizures and for sudden unexpected death in epilepsy (SUDEP), and many nocturnal seizures remain undetected in unattended patients. A portable automatic seizure detection device will be an important tool for helping these patients.