GLEAM: Noninvasive Glucose Measurement Using Impedance Tomography (GLEAM)

GLEAM: Noninvasive Glucose Measurement Using Impedance Tomography - a Pilot Project

The GLEAM study aims at assessing the potential of electrical impedance tomography (EIT) for noninvasive glucose measurement.

Study Overview

• Status: Completed
• Conditions: Diabetes Mellitus
• Intervention / Treatment: Other: Controlled euglycemia, hypoglycemia and hyperglycemia

Detailed Description

Within the GLEAM study, paired samples of EIT and blood glucose measurements will be collected in individuals with type 1 diabetes during standardized euglycemia, hypoglycemia and hyperglycemia. These samples will be used to assess the potential of EIT for noninvasive glucose measurement and/or dysglycemia detection.

• Study Type: Interventional
• Enrollment (Actual): 16
• Phase: Not Applicable

Contacts and Locations

Study Locations

• Switzerland
  • Bern, Switzerland
    • Department of Diabetes, Endocrinology, Nutritional Medicine and Metabolism

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult
• Accepts Healthy Volunteers: No

Description

Inclusion Criteria:

• Written, informed consent
• Type 1 Diabetes mellitus as defined by WHO for at least 6 months
• Aged 18 - 60 years
• HbA1c ≤ 9.0 %
• Insulin treatment with good knowledge of insulin self-management
• Use of a continuous (CGM) or flash glucose monitoring system (FGM)
• Native language German or Swiss German

Exclusion Criteria:

• Incapacity to give informed consent
• Contraindications to insulin aspart (NovoRapid®)
• Known allergies to adhesives of the EIT device (e.g., gel electrodes)
• Pregnancy, breast-feeding or lack of safe contraception
• Active heart, lung, liver, gastrointestinal, renal or psychiatric disease
• Patients with implantable electronic devices (e.g., pacemaker or implantable cardioverter defibrillator (ICD)) or thoracic metal implants
• Epilepsy or history of seizure
• Active drug or alcohol abuse
• Chronic neurological or ear-nose-and-throat (ENT) disease influencing voice or history of voice disorder
• Thoracic or back deformities
• Body mass index (BMI) >35.0 kg/m2
• Open wounds, burns, or rashes on the upper thorax
• Active smoking
• Medication known to interfere with voice or to induce listlessness (e.g., opioids, benzodiazepines, etc.)

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Other
• Allocation: N/A
• Interventional Model: Single Group Assignment
• Masking: None (Open Label)

Number of Arms

1

Arms and Interventions

Participant Group / Arm	Intervention / Treatment
Other: Controlled euglycemia, hypoglycemia and hyperglycemia	Other: Controlled euglycemia, hypoglycemia and hyperglycemia; EIT measurements are collected in different glycemic states (euglycemia, hypoglycemia and hyperglycemia). Venous blood glucose is measured using a gold-standard glucose analyzer.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Change of the electrical impedance tomography (EIT) signal of the thoracic region across the glycemic trajectory.	EIT signals will be collected at multiple frequencies between 50 kHz and 1 MHz from the thoracic region in euglycemia, hypoglycemia and hyperglycemia using a multi-channel EIT measurement device.	5 hours

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Change of hypoglycemia symptoms across the glycemic trajectory.	Hypoglycemia symptoms will be collected in euglycemia, hypoglycemia and hyperglycemia using a standardized questionnaire (Edinburgh Hypoglycemia Scale, a higher score means more symptoms, minimum score 7 points, maximum score 77 points).	5 hours
Voice parameters indicative of dysglycemia	Voice data will be collected using a microphone in euglycemia, hypoglycemia and hyperglycemia. After sampling, an interpretable machine learning (ML) method will be used to identify voice parameters indicative of dysglycemia.	5 hours
Change in cognitive performance across the glycemic trajectory.	Cognitive performance will be assessed using the Trail Making B Test (more time needed to complete the tests means worse cognitive performance).	5 hours
Change in cognitive performance across the glycemic trajectory.	Cognitive performance will be assessed using the Digit Symbol Substitution Test (higher score means better cognitive performance).	5 hours
Performance of a machine learning model to detect dysglycemia from the above-mentioned signals (EIT, symptoms, voice, physiological signals) quantified as area under the receiver operating characteristics curve (AUROC).	Signals for machine learning modeling will be collected in euglycemia, hypoglycemia and hyperglycemia.	5 hours
Performance of a machine learning model to detect dysglycemia from the above-mentioned signals (EIT, symptoms, voice, physiological signals) quantified as sensitivity.	Signals for machine learning modeling will be collected in euglycemia, hypoglycemia and hyperglycemia.	5 hours
Performance of a machine learning model to detect dysglycemia from the above-mentioned signals (EIT, symptoms, voice, physiological signals) quantified as specificity.	Signals for machine learning modeling will be collected in euglycemia, hypoglycemia and hyperglycemia.	5 hours
Performance of the machine learning model to predict glucose values from the above-mentioned signals (EIT, symptoms, voice, physiological signals) quantified as root mean squared error (RMSE).	Signals for machine learning modeling will be collected in euglycemia, hypoglycemia and hyperglycemia.	5 hours
Performance of the machine learning model to predict glucose values from the above-mentioned signals (EIT, symptoms, voice, physiological signals) quantified as mean absolute relative difference (MARD).	Signals for machine learning modeling will be collected in euglycemia, hypoglycemia and hyperglycemia.	5 hours
Performance of the machine learning model to predict glucose values from the above-mentioned signals (EIT, symptoms, voice, physiological signals) using Bland-Altman plots.	Signals for machine learning modeling will be collected in euglycemia, hypoglycemia and hyperglycemia.	5 hours
Performance of the machine learning model to predict glucose values from the above-mentioned signals (EIT, symptoms, voice, physiological signals) using the Clarke Error Grid.	Signals for machine learning modeling will be collected in euglycemia, hypoglycemia and hyperglycemia.	5 hours

Collaborators and Investigators

• Sponsor: Insel Gruppe AG, University Hospital Bern
• Collaborators: Idiap Research Institute; CSEM Centre Suisse d'Electronique et de Microtechnique SA
• Investigators: Principal Investigator: Christoph Stettler, Prof. MD, Department of Diabetes, Endocrinology, Nutritional Medicine and Metabolism; Bern, Switzerland

Study record dates

Study Major Dates

• Study Start (Actual): January 31, 2024
• Primary Completion (Actual): April 10, 2024
• Study Completion (Actual): April 10, 2024

Study Registration Dates

• First Submitted: December 19, 2023
• First Submitted That Met QC Criteria: January 15, 2024
• First Posted (Actual): January 25, 2024

Study Record Updates

• Last Update Posted (Actual): May 2, 2024
• Last Update Submitted That Met QC Criteria: April 30, 2024
• Last Verified: April 1, 2024

More Information

Terms related to this study

• Other Study ID Numbers: GLEAM

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No