Prevention of Severe Hypoglycemia by Continuous EEG Monitoring

Background: The brain is dependent on constant glucose supply and hypoglycaemia results in reduced cognition, unconsciousness, seizures, and possible death. Prevention of hypoglycaemia is accordingly a key point in diabetes treatment. The effect of hypoglycemia on the electrical activity of the brain is well described. We propose an alarm device for severe hypoglycaemia based on continuous electroencephalography (EEG), real-time data analysis by an automated algorithm and an auditory alarm. Methods and results: People with type 1 diabetes (T1D) were exposed to hypoglycaemia by excess insulin administration. EEG was scored visually by neurophysiologists for deviations compatible with neuroglycopenia. From these initial experiments, a multiparameter algorithm was developed by applying an artificial neural network. Subsequently series of experiments were conducted in T1D patients in order to improve the algorithm and to test its clinical applicability. Thus people with T1D with normal, reduced, or absent awareness of hypoglycaemia were exposed to hypoglycaemia both during daytime and during sleep. EEG was analyzed by the automated algorithm. All patients developed hypoglycaemia-associated EEG changes. During the sleep experiments, these changes occurred irrespective of sleep stage, and the majority of patients woke up when they received an auditory alarm at the onset of hypoglycaemia-associated EEG changes. We found that the EEG changes were independent of diabetes duration, awareness status, and counter-regulatory hormone response. In subsequent studies, a miniaturized partly implanted EEG recorder was tested. The device consists of a small implant recording the EEG and an external device which wirelessly receives EEG and gives an alarm when hypoglycaemia-associated EEG changes are apparent. In preliminary studies of one month duration, we found that the device was well tolerated and successfully warned the patients in time to take appropriate action before severe hypoglycaemia was present. Conclusion: The results obtained so far hold promises for the development of an EEG-based alarm for severe hypoglycaemia in people with T1D. Hypoglycaemia-associated EEG changes seem to be a general feature which makes a common algorithm applicable. Further studies are needed to define the sensitivity and tolerability of the partly implanted EEG-based device for severe hypoglycaemia during long-term use.