The aim of this study is based on recent pilot studies carried out at Odense University Hospital showing that the acute changes in electroencephalographic (EEG) signals (i.e. electrical activity inthe brain) elicited by insulin-induced hypoglycemia in patients with type 1 diabetes can be reliable detected by real-time processing of these EEG signals using mathematical algorithms and state of the art noise and artifact reduction. These preliminary results also showed that the hypoglycemia-induced EEG changes are detectable 15-30 min before deterioration in cognitive function impedes an adequate response to warning. We hypothesize that these observations apply to the majority of patients with type 1 diabetes, and therefore, that it is possible to develop an automated device to detect hypoglycemic episodes by continuous real-time monitoring and processing of EEG signals. To test our hypothesis, the specific aims of the present proposal are:

1. Detection of hypoglycemia-induced EEG changes using subcutaneous electrodes
2. Ambulatory EEG monitoring using subcutaneous electrodes

The near-normalization of glycemic control has become an established treatment goal in diabetes in order to reduce late complications such as nephropathy, neuropathy, retinopathy and cardiovascular disease (1,2). However, the frequency of insulin-induced hypoglycemia increases several-fold during intensified insulin therapy (2,3). Thus, hypoglycemia is the most common acute complication in the treatment of diabetes with insulin. During hypoglycemia the cognitive function is disturbed, and may progress to unconsciousness and seizures. This can lead to high-risk situations, e.g. while driving or operating a machine. Estimates of deaths in patients with type 1 diabetes attributed to hypoglycemia vary between 2% and 6% (4,5). Moreover, the risk of hypoglycemia limits everyday activities of diabetic patients decreasing their quality of life. It is therefore not surprising that hypoglycemia is the most feared acute complication of insulin therapy in diabetic patients. This fear of hypoglycemia discourages diabetic subjects from attempting to maintain tight glycemic control, which in turn leads to a higher incidence of late complications and consequently increased mortality rate (1,6,7) In the first years of type 1 diabetes, most patients are able to sense the characteristic symptoms of hypoglycemia, which can then be relieved by consuming appropriate food. The symptoms of hypoglycemia can roughly be classified as autonomic (warning) symptoms caused by the release of catecholamines, and neuroglycopenic symptoms caused by the lack of glucose in the brain. In many patients symptoms are often compromised at night (nocturnal asymptomatic hypoglycemia) due to impaired glucose counterregulatory response by adrenaline and glucagon. The chronic form of hypoglycemia unawareness is very common. A quarter of all insulin-treated diabetic patients have some degree of diminished symptomatic awareness, but this proportion increases to almost 50% in patients who have had diabetes for more than 20 years (8). Strict control of diabetes by intensive insulin therapy is associated with increased risk of the hypoglycemia unawareness syndrome with loss of autonomic warning symptoms (2,6,9). This seems to involve diminished hormonal glucose counterregulation due to recurrent hypoglycemic episodes (6,9).

For these reasons, a number of studies have been carried out with the aim of developing automatic detection systems, which can warn the diabetic patients before blood glucose levels are reduced to the level at which severe neuroglycopenia develops, typically about 2.0-2.5 mmol/l. Most studies have evaluated the potential of continuous glucose monitoring (CMG) to decrease the frequency of hypoglycemia. Although, smaller studies have reported a lower risk of hypoglycemia using CMG compared with conventional glucose measurements (10,11), larger multi-center studies have failed to reproduce these findings (11-14). This could be explained by a low accuracy in the low range of glucose values and delay in detection time during rapid changes using CMG (11,13,14). In fact, CMG only recognizes less than 50% of hypoglycemic events (15). Thus, even with a marginal improvement compared with conventional glucose measurements, CMG is far from the goal of completely avoiding severe hypoglycemic episodes.

The EEG signal reflects the functional state and metabolism of the brain. The brain is almost totally dependent on a continuous supply of glucose, and when this is lower than the metabolic requirements of the brain, its function deteriorates. Indeed, neuroglycopenic hypoglycemia in insulin-treated diabetic patients is associated with characteristic changes in EEG with a decrease in alpha activity, an increase in delta activity, and in particular an increase in theta activity (16-19). These changes are clearly seen at ~2.0 mmol/l (16,17), but may be present already at higher glucose levels (~3.0 mmol/l), in particular in type 1 diabetic patients with hypoglycemic unawareness (19,20). It has been shown that the most characteristic changes, the increase in theta activity, appears 19 min before severe cognitive impairment (20). This suggests a "window" between hypoglycemia-induced EEG changes and severe neuroglycepenia, which is an important prerequisite in developing an automatic detection system capable of warning the patient.

A number of studies have characterized the changes in the EEG that results from hypoglycemia (16-20), but none have proposed a method of processing and testing in real-time. With a device, which can perform real-time monitoring and processing of EEG signals and automatically detect and warn the patient of hypoglycemia-induced EEG changes, it would be possible for the patient to avoid severe neuroglypenic symptoms e.g. by ingestion of carbohydrates. The construction of an EEG-based hypoglycemia alarm system must fulfill the following criteria. First, the device should be able to distinguish hypoglycemia-induced EEG changes from normal changes in EEG, noise and artifacts with high sensitivity and specificity using a mathematical algorithm that classifies the EEG in real-time. Second, these EEG changes should be observed in the majority of insulin-treated diabetic patients during hypoglycemia. Third, there should be a "window" between hypoglycemia-induced EEG changes and severe cognitive impairment. Moreover, the device should be fully compatible with normal everyday activities. Thus, the electrodes should be thin and implanted subcutaneously, and the monitoring and processing unit should be small, have sufficient battery power, and capable of communicating with a PDA or cell phone.

Inclusion Criteria

• 18-60 year old subjects
• Type 1 diabetics with complete or partial hypoglycemia unawareness.
• Ability to comprehend and a willingness to sign an informed consent form

Exclusion Criteria:

• Neurological or psychiatric disease.
• Current use of neuroactive medication or recreational drugs.
• Pregnancy.
• Patients with known heart disease, former myocardial infarction or cardiac arrhythmia
• Patients with known epilepsy or in treatment with anti-epileptic drugs for all purposes
• Patients treated with drugs that are known to influence the EEG, including benzodiazepines and other anxiolytics, anti-depressants and beta-blocking agents
• Patients that are judged incapable of understanding the patient information or who are not capable of carrying through the investigation
• Cancer of any kind

Twenty adult patients with type 1 diabetes will be participating in the study.