Logo: to the web site of Uppsala University

Introduction: Hypoglycemia in type 1 diabetes (T1D) ranges from mild to life-threatening events, yet most studies of hypoglycemia frequency rely on self-reported or aggregated data. Residual endogenous insulin production is associated to fewer severe hypoglycemic events, highlighting the potential benefit of preserving or restoring insulin production. For this purpose, gamma-aminobutyric acid (GABA) has emerged from experimental studies as a potential therapeutic drug candidate.

Aim: This thesis aimed to investigate the real-world frequency of hypoglycemia in children and adolescents with T1D, and to evaluate GABA’s therapeutic potential in a clinical trial.

Methods: Five studies were included. Endogenous GABA, C-peptide, counter-regulatory hormones and cytokine levels were analyzed in plasma. A controlled-release oral formulation of GABA (Remygen®) was assessed in a randomized controlled Phase I/II clinical trial in individuals with long-standing T1D (n=35) for safety, effect on endogenous insulin production and hypoglycemic counter-regulation.

The real-world frequency of hypoglycemia and its relationship to overall metabolic control and age was evaluated using retrospective continuous glucose monitoring (CGM)-data and clinical records. More than 50,000 hypoglycemic events were analyzed. Additionally, a single-metric scoring model for CGM-data evaluation was developed based on n=82,114 days of CGM-data by assessing three dimensions of glucose control. The models validity was evaluated against clinical treatment targets and interpretations of a clinical expert board (CEB). 

Results: GABA levels did not differ between individuals with T1D and healthy controls, but correlated with anti-GAD and cytokines. GABA treatment showed no improvements in endogenous insulin production or hypoglycemic counter-regulation, but side-effects were commonly observed. In the retrospective studies on CGM-data, mild hypoglycemic events (<3.9 mmol/L) were common. On average occurring on a near daily basis, regardless of age or metabolic control. However, no increased risk of severe- or serious (<3.0 mmol/L) hypoglycemia was observed in children achieving HbA1c ≤48 mmol/mol. The developed CGM scoring model correlated well with CGM-metrics and CEB interpretations.

Conclusions: Despite technological advancements, hypoglycemia remains a persistent challenge in T1D. GABA failed to regain beta-cell function, underscoring the need for alternative therapies in this aspect. Meanwhile, models for enhanced CGM analyses may aid in optimizing glucose management.

Gamma-amino butyric acid (GABA) has in experimental studies been found to promote beta-cell proliferation, enhance insulin secretion and reduce inflammation, positioning it as a candidate drug for type 1 diabetes (T1D) therapy. This phase I/II randomized controlled trial assessed the safety and efficacy of long-term treatment with Remygen® (Diamyd Medical), a controlled-release oral GABA formulation, as a potential beta-cell regenerative therapy in adults with long-standing T1D. Thirty-five male subjects with T1D (≥ 5 years) were randomized into three arms receiving the study drug(s) once daily for 6 months: GABA 200 mg (Arm 1), GABA 600 mg (Arm 2) and GABA 600 mg + alprazolam 0.5 mg for 3 months followed by GABA 600 mg alone for 3 months (Arm 3). Safety measures, hormonal counter-regulation during hypoglycemic clamps, fasting- and stimulated C-peptide levels, were assessed at multiple timepoints. Safety concerns included elevated aspartate aminotransferase (AST) in nine subjects, leading to the withdrawal of two subjects. Most elevations were, however, transient with no dose-differences. No effects were observed on fasting- or stimulated C-peptide levels, CGM metrics or HbA1c. Hypoglycemic hormonal counter-regulation was unaltered. To conclude, we found no clinical evidence of a beta-cell regenerative effect of GABA, but side effects were commonly observed.

Introduction Despite the improvements in diabetes management by continuous glucose monitoring (CGM) it is difficult to capture the complexity of CGM data in one metric. We aimed to develop a clinically relevant multidimensional scoring model with the capacity to identify the most alarming CGM episodes and/or patients from a large cohort.Research design and methods Retrospective CGM data from 2017 to 2020 available in electronic medical records were collected from n=613 individuals with type 1 diabetes (total 82 114 days). A scoring model was developed based on three metrics; glycemic variability percentage, low blood glucose index and high blood glucose index. Values for each dimension were normalized to a numeric score between 0-100. To identify the most representative score for an extended time period, multiple ways to combine the mean score of each dimension were evaluated. Correlations of the scoring model with CGM metrics were computed. The scoring model was compared with interpretations of a clinical expert board (CEB).Results The dimension of hypoglycemia must be weighted to be representative, whereas the other two can be represented by their overall mean. The scoring model correlated well with established CGM metrics. Applying a score of >= 80 as the cut-off for identifying time periods with a 'true' target fulfillment (ie, reaching all targets for CGM metrics) resulted in an accuracy of 93.4% and a specificity of 97.1%. The accuracy of the scoring model when compared with the CEB was high for identifying the most alarming CGM curves within each dimension of glucose control (overall 86.5%).Conclusions Our scoring model captures the complexity of CGM data and can identify both the most alarming dimension of glycemia and the individuals in most urgent need of assistance. This could become a valuable tool for population management at diabetes clinics to enable healthcare providers to stratify care to the patients in greatest need of clinical attention.

Introduction

To improve the utilization of continuous- and flash glucose monitoring (CGM/FGM) data we have tested the hypothesis that a machine learning (ML) model can be trained to identify the most likely root causes for hypoglycemic events.

Methods

CGM/FGM data were collected from 449 patients with type 1 diabetes. Of the 42,120 identified hypoglycemic events, 5041 were randomly selected for classification by two clinicians. Three causes of hypoglycemia were deemed possible to interpret and later validate by insulin and carbohydrate recordings: (1) overestimated bolus (27%), (2) overcorrection of hyperglycemia (29%) and (3) excessive basal insulin presure (44%). The dataset was split into a training (n = 4026 events, 304 patients) and an internal validation dataset (n = 1015 events, 145 patients). A number of ML model architectures were applied and evaluated. A separate dataset was generated from 22 patients (13 ‘known’ and 9 ‘unknown’) with insulin and carbohydrate recordings. Hypoglycemic events from this dataset were also interpreted by five clinicians independently.

Results

Of the evaluated ML models, a purpose-built convolutional neural network (HypoCNN) performed best. Masking the time series, adding time features and using class weights improved the performance of this model, resulting in an average area under the curve (AUC) of 0.921 in the original train/test split. In the dataset validated by insulin and carbohydrate recordings (n = 435 events), i.e. ‘ground truth,’ our HypoCNN model achieved an AUC of 0.917.

Conclusions

The findings support the notion that ML models can be trained to interpret CGM/FGM data. Our HypoCNN model provides a robust and accurate method to identify root causes of hypoglycemic events.

Introduction: Hypoglycemia composes an always present risk in the treatment of type 1 diabetes (T1D) and can be a fatal complication. Many studies on hypoglycemic events are based on self-reported data or focused on the aggregated time below range. We have processed continuous glucose monitoring (CGM) data in children and adolescents with T1D in order to examine all occurring hypoglycemic events.

Research design and methods: CGM data (mean 168 +/- 3 days) from 214 children and adolescents with T1D were analyzed using computer-based algorithms. Patients were divided into three groups based on estimated HbA1c (eHbA1c): (1) <= 48 mmol/mol (n=58); (2) 49-64 mmol/ mol (n=113); (3) >= 65 mmol/mol (n=43). The groups were compared concerning descriptive data and CGM metrics with emphasis on the frequency of hypoglycemic events.

Results: Only one self-reported event of severe hypoglycemia was registered, while 54 390 hypoglycemic events (<3.9 mmol/L (<70 mg/dL)) were identified from CGM data out of which 11 740 were serious (<3.0 mmol/L (<54 mg/dL)). On average there were 1.5 +/- 0.1 hypoglycemic events per 24 hours out of which 1.2 +/- 0.1 were mild (3.0-3.9 mmol/L) and 0.3 +/- 0.02 serious. Group 1 had a higher frequency of both total and mild hypoglycemic events compared with both groups 2 and 3. However, the frequency of serious hypoglycemic events was similar in all groups. A negative correlation was observed for eHbA1c and total daily and mild hypoglycemic events (r=-0.57 and r=-0.66, respectively, p<0.0001), whereas for serious hypoglycemic events there was only a borderline significance (r=-0.13, p=0.05).

Conclusions: This study shows that hypoglycemic events are a frequent phenomenon in children and adolescents with T1D, occurring regardless of overall metabolic control. Although patients with an HbA1c =48 mmol/mol had a higher frequency of mild hypoglycemic events there was no increase in serious hypoglycemic events.

Gamma-aminobutyric acid (GABA) is an important inhibitory neurotransmitter in the central nervous system (CNS) and outside of the CNS, found in the highest concentrations in immune cells and pancreatic beta-cells. GABA is gaining increasing interest in diabetes research due to its immune-modulatory and beta-cell stimulatory effects and is a highly interesting drug candidate for the treatment of type 1 diabetes (T1D). GABA is synthesized from glutamate by glutamic acid decarboxylase (GAD), one of the targets for autoantibodies linked to T1D. Using mass spectrometry, we have quantified the endogenous circulating levels of GABA in patients with new-onset and long-standing T1D and found that the levels are unaltered when compared to healthy controls, i.e., T1D patients do not have a deficit of systemic GABA levels. In T1D, GABA levels were negatively correlated with IL-1 beta, IL-12, and IL-15 15 and positively correlated to levels of IL-36 beta and IL-37. Interestingly, GABA levels were also correlated to the levels of GAD-autoantibodies. The unaltered levels of GABA in T1D patients suggest that the GABA secretion from beta-cells only has a minor impact on the circulating systemic levels. However, the local levels of GABA could be altered within pancreatic islets in the presence of GAD-autoantibodies.