PHILADELPHIA – Reliable, externally worn seizure-detection devices for epilepsy patients have been eagerly sought but elusive. New results from the pivotal trial of an arm-mounted device that works by analyzing skin electromyography showed promise when tested in 142 adults and children with epilepsy who also underwent video EEG monitoring, Dr. José E. Cavazos reported in a poster at the annual meeting of the American Epilepsy Society.
More than 7,000 hours of simultaneous surface electromyography with the investigational device, worn on the patient’s upper arm, and video EEG monitoring as the gold standard for seizure detection showed that the device correctly identified all 29 generalized tonic-clonic seizures experienced by the 142 patients studied. This level of detection came with a false-positive rate of 1.47 false events during every 24 hours of monitoring, with most of the false positives occurring when the device was being placed or removed.
The results made Dr. Cavazos hopeful that the Food and Drug Administration would clear the device, the Brain Sentinel Seizure Detection and Warning System, for U.S. marketing in 2016. The FDA first received the marketing-clearance application in late 2014, said Dr. Cavazos, an epileptologist and professor of neurology at the University of Texas Health Science Center in San Antonio.
But Dr. Cavazos also conceded that receiving FDA clearance for U.S. sale of the device is just the next step in an ongoing process to prove its clinical value.
“Currently, clinical decisions for epilepsy patients are often based on patient-reported seizure counts. There is a clear need for objective data on seizure frequency,” something now obtainable only by running an expensive visual-monitoring study, said Dr. Cavazos, who is also founder of and a stockholder in the company that is developing this device. “We definitely need to show in future studies that monitoring [with the worn device] has a positive impact on patient care, by reducing hospitalizations and producing improved patient outcomes,” he said in an interview. “These data may be enough to get regulatory clearance to sell the device, but insurers may not want to reimburse for it until we can prove its value.”
When the device detects a generalized tonic-clonic seizure, it produces an audible alert and also starts recording both surface EMG data and local audible sounds for use in future analysis of the patient’s status. In the study, seizure detection occurred an average of 14 seconds following seizure onset as judged by visual EEG monitoring adjudicated by three epileptologists. The study also included an additional 31 patients who initially entered the study and began wearing the device but did not complete the monitoring protocol because of an adverse effect while wearing the monitor, usually mild or moderate skin irritation.
The results from this study come in a context of extensive research efforts aimed trying to find reliable and unobtrusive methods to quickly detect seizures short of running an observational study. The best results so far have come when using devices that monitor brain EEG signals with electrode-containing caps or implanted electrodes. But patients have voiced substantially reduced interest in having to resort to such stigmatizing methods, noted Dr. Elizabeth Donner during a separate talk at the meeting.
“Patients don’t want what looks like an EEG-detection device. We need devices that suit our patients’ needs,” said Dr. Donner, an epileptologist at the Hospital for Sick Children in Toronto. Aside from direct measurement of EEG, other approaches involve measurement of heart rate or movement, she noted.
“There are very good techniques for detecting seizures without using EEG,” said Dr. Gregory L. Krauss, a professor of neurology at Johns Hopkins University in Baltimore who joined Dr. Donner in a discussion of the pros and cons of current methods for seizure detection. He specifically cited a heart-rate based detection scheme reported last year by a Belgian team (Seizure. 2015 Nov;:52-61). But like all detection methods, the heart rate–based approach showed a marked trade-off between sensitivity and specificity: A heart-rate threshold of at least 50% above baseline had a sensitivity of 82% while identifying one false positive every hour. Raising the detection threshold to at least 30% above baseline boosted the sensitivity to 91%, but with the cost of flagging 3.5 false positives each hour.
Dr. Krauss is currently trying to develop a motion-based approach to identifying seizure onset using an Apple watch, He concluded that, as of now, development of seizure-detection devices that do not rely on direct measurement of brain EEG patterns remains a work in progress, with no prospectively collected data, although he made this assessment prior to the poster report by Dr. Cavazos at the meeting.