Seizure detection devices for use in antiseizure medication clinical trials: A systematic review
Introduction from Dr Markus Reuber, editor-in-chief of Seizure
Given that adults with epilepsy undergoing video-EEG monitoring fail to report over 85% of seizures in sleep and up to 50% of those occurring in the waking state, it should be a matter of major concern that the development and licencing of antiepileptic drugs (AEDs) and anti-seizure devices continues to rely on patients’ self report of the number and frequency of their seizures. The fact that pre-licensing studies employ randomised controlled trial (RCT) designs does not address this concern: It is entirely feasible that a drug under investigation could have a greater effect on patients becoming aware of or remembering their seizures without having any greater impact on seizures frequency or severity than placebo.
The use of wearable devices capable of detecting seizures seems to be an obvious answer to the deficiencies of patients’ self report. There has been impressive progress with the development of such devices over recent years using a range of different measures to identify seizures. This progress has been reflected by a two studies published in Seizure over the last year – one demonstrating the potential of ear-EEG based EMG analysis (2) and another using accelerometry-based sensors to detect bilateral tonic clonic seizures (3).
My Editor’s Choice from the current issue of Seizure is a review article by Abhinav Kurada et al. which focuses on this idea and explores how close we have come to employing wearable automatic detection devices in AED development studies (4). This review is based on an analysis of 38 original research papers about commercial or non-commercial devices providing sufficient data to allow the calculation of an F1-score based on positive (correct) detections, false positive detections and false negative detections.
The review identified a number of devices capable of outperforming patient report in relation to bilateral tonic seizure and focal seizures with impaired awareness seizures. Devices were also capable of capturing potentially useful secondary outcome measures such as seizure duration and vital signs during the ictal period. However, so far none of the reviewed devices have been shown to be capable of reliably detecting focal seizures with retained awareness, atonic or clonic seizures. What is more, many of the device validations studies had significant shortcomings such as a small sample size, the inclusion of highly specific patient populations, short study duration and a lack of replication across multiple patient cohorts. Another problem is that (for obvious methodological reasons) most validation studies were carried out in Epilepsy Monitoring Unit settings. This means that participants were restricted in the range of activities they could engage in during monitoring. The performance of different devices was difficult to compare because studies had employed a range of different validation measures. This means that, although devices hold much promise and are highly likely to complement or replace patient reported outcome measures in the future, there is more work to be done. Kurada et al. propose a logical framework for this work involving sequential inpatient validation, outpatient validation, and experimental validation in clinical intervention trials.
(1) Hoppe C, Poepel A, Elger CE. Epilepsy: accuracy of patient seizure counts. Arch Neurol 2007;64(11):1595–9.
(2) Zibrandtsen IC et al. Detection of generalized tonic-clonic seizures from ear-EEG based on EMG analysis. Seizure 2018;59:54-59.
(3) Johannson D et al. Tonic-clonic seizure detection using accelerometry-based wearable sensors: a prospective, video-EEG controlled study. Seizure 2018;65:48-54.
(4) Kurada AV, Srinivasan T, Hammond S, Ulate-Campos A, Bidwell J. Detection Devices for use in Antiseizure Medication Clinical Trials: A Systematic Review. Seizure 2019;66:61-69.