Abstracts

Rationale: Individuals with epilepsy have a 1.9-9.3 higher mortality risk than those without this condition. Predicting seizure onset could reduce morbidity and mortality, and substantially improve quality of life in persons with epilepsy and their caregivers. Seizure-alert dogs could accomplish this by warning the patient of impending seizure. One study demonstrated that trained dogs can detect odors of seizures (Catala, et al., 2019). It remains to be shown if dogs also could detect the pre-ictal state. We evaluated a dog’s ability to detect pre-ictal odor where odor samples were obtained in a controlled video-EEG hospital setting and a seizure-alert dog was tested in a separate, controlled environment with multiple relevant distractors.

Methods: Patients 18 years and older already being admitted to a single-center epilepsy monitoring unit for video-EEG monitoring for seizure characterization were asked to participate. Odor samples were obtained using cotton balls swiped on the neck and hand and exhaled on by the patient. Samples were obtained regarding the first seizure (ictus) and all subsequent seizures. After the first seizure, the patient began obtaining cotton ball samples every 15 minutes while awake with each sample being time-stamped. When a second or other subsequent seizures occurred, the recent pre-ictal samples were saved. Other distractor/control samples were collected at admission, discharge, and after 15 minutes of exercise. Samples were sent in odor proof bags to the Medical Mutts, Inc facility for testing. A seizure-alert dog previously trained on other samples was let into a testing room alone to choose 1 “hot” (pre-ictal) sample from 6 distractors placed in identical metal receptables. If the dog identified the correct sample, as determined by video-monitoring from another room, the dog was given a treat by a “pet tutor” device and results notated.

Results: We completed pilot testing of our protocol in 5 participants who had epileptic seizures observed and confirmed with concurrent video-EEG by a neurologist. Samples were interpreted by 1 dog. There were 10 ictal samples, 19 pre-ictal samples, and 174 distractor samples. Compared to chance prediction of 14% (1 of 7 samples were “hot”), the dog correctly identified 9 of 10 ictal samples (95% Confidence Interval 70-100%). The dog identified 19 of 19 pre-ictal samples (95% Confidence Interval 89-100%) ranging from 3 minutes to 75 minutes prior to each seizure. In one trial with no “hot” ictal or pre-ictal sample, the dog successfully did not alert.

Conclusions: This seizure-alert dog correctly identified all but one seizure or pre-ictal sample where the dog was alone during identification and seizures were confirmed with video-EEG. This rigorous study design improved upon previous studies of seizure-alert dogs and further demonstrates the abilities of seizure-alert dogs to identify both seizure (ictal) and pre-ictal periods. Positive identification of pre-ictal samples up to 75 minutes before seizure prompts further studies to identify how long before seizures that seizure-alert dogs can predict the impending seizure.

Funding: Cottage Health Research Institute grant