Diabetes Type 1
Conditions
Brief summary
Use of devices for continuous monitoring of the blood sugar is valuable for people with diabetes to understand their disease and to help prevent low blood sugar. Furthermore, continuous monitoring should be used in drug development to evaluate efficacy and safety. However, the devices have been criticised for being too inaccurate. This investigation sought to reveal the inaccuracies of current devices and to assess the subsequent usability related to the mentioned use cases.
Detailed description
The following study is an exploratory investigation of continuous glucose monitoring based on data from a completed Novo Nordisk A/S clinical trial. Please refer to ClinicalTrials.gov Identifier: NCT02825251. Continuous Glucose Monitoring (CGM) provides an interstitial glucose reading every 5 minutes and is thus a powerful and important tool to identify glycaemic variability in people with diabetes. CGM is valuable for people with diabetes to understand their glucose metabolism and it has the potential to be used for detection and prediction of glycaemic excursions, such as, the potentially fatal and inevitable events of hypoglycaemia, or even as a component in the holy grail of diabetes technology; the artificial pancreas. However, CGM has been criticised for being inaccurate and unreliable, amongst others, due to the physiological and a device-related delay between plasma glucose (PG) and interstitial glucose (IG). Nevertheless, CGM keeps on being popular and in February 2017 an international consensus was established at the Advanced Technologies & Treatments for Diabetes (ATTD) congress that even considers CGM data as a valuable and meaningful end point to be used in clinical trials of new drugs and devices for diabetes treatment where accuracy is of high importance. The above mentioned use cases entail that the CGM data are accurate. Therefore, the first part of this research proposal is to investigate whether the newest state-of-the-art CGM devices used in Novo Nordisk trials are in fact accurate. Based on these results, it is investigated to which degree glycaemic variability can be revealed. To investigate the accuracy of CGM, mean absolute relative difference (MARD) will be calculated and presented and the impact of the delay assessed by time shifting CGM measurements. Furthermore, correlation analyses, between for example, PG and first derivative of IG, will be performed to try to understand when CGM devices tend to measure inaccurate. Lastly, machine learning and/or deep learning approaches will be utilised to reveal glycaemic patterns and to detect/predict outcomes, such as, hypoglycaemia. Different glycaemic variability investigations will be undertaken: * Test of PG vs IG and effect on clinical research. \[analysis of differences\] * Correlation between PG values at bedtime and nocturnal hypoglycaemic events \[correlation analyses\] * Effect of main evening meal and meal-time dose on nocturnal hypoglycaemic events \[correlation analyses\] * Prediction of PG-confirmed hypoglycaemic events with CGM, dose and meal data as input \[machine learning\] * The optimal dose and meal distribution and least CGM variability / eHbA1c \[machine learning\] * Algorithm to suggest optimal dosing in relation to glycaemic variability \[machine learning\] Requested data are demographic, CGM, meal, dose and hypoglycaemia data from the following trial. The analyses are independent of treatment and therefore the treatment arm can be blinded.
Interventions
OTHERCGM
This study seeks to assess CGM accuracy and develop prediction models for hypoglycemia detection and no intervention is therefore applied.
Sponsors
Peter Vestergaard
Study design
Observational model
OTHER
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Inclusion criteria
As copied from the original clinical trial ClinicalTrials.gov Identifier: NCT02825251 Inclusion Criteria: * Male or female, age at least 18 years at the time of signing the informed consent * Diagnosed with T1DM (Type 1 Diabetes Mellitus) (based on clinical judgement and/or supported by laboratory analysis as per local guidelines) equal or above 1 year prior to the day of screening * Using the same Medtronic pump (Minimed 530G (551/751), Paradigm Veo (554/754), Paradigm Revel (523/723), Paradigm (522/722)) for CSII in a basal-bolus regimen with a rapid acting insulin analogue for at least six months prior to screening and willing to stay on the same pump model throughout the trial (if the model is changed the change should not exceed 7 consecutive days.) * HbA1c (glycosylated haemoglobin) 7.0-9.0% (53-75 mmol/mol) as assessed by central laboratory at screening * Body mass index (BMI) below or equal to 35.0 kg/m\^2 at screening * Ability and willingness to take at least 3 daily meal-time insulin bolus infusions every day throughout the trial
Exclusion criteria
* Any of the following: myocardial infarction, stroke, hospitalization for unstable angina or transient ischaemic attack within the past 180 days prior to the day of screening * Planned coronary, carotid or peripheral artery revascularisation known on the day of screening * History of hospitalization for ketoacidosis below or equal to 180 days prior to the day of screening * Treatment with any medication for the indication of diabetes or obesity other than stated in the inclusion criteria in a period of 90 days before screening * Any condition which, in the opinion of the Investigator, might jeopardise a Subject's safety or compliance with the protocol
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Optimal Time Shift of Continuous Glucose Monitoring Measurements | 16 weeks | Continuous glucose monitoring (CGM) measurements are delayed compared to blood glucose. The CGM signal is time-shifted -1 minute at a time and the mean absolute difference between CGM and blood glucose measurements are calculated at each step. The lowest mean absolute difference depicts the optimal time shift in minutes. The resultant mean absolute relative difference is provided as outcome. Publication reference: https://doi.org/10.1177/1932296819848721 |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Area Under the Receiver Operating Characteristics Curve of the Hypoglycemia Prediction | 16 weeks | Area under the receiver operating characteristics curve (ROC-AUC) is a measure of the prediction capabilities of a prediction algorithm. Each point of the curve gives a sensitivity and a specificity of the prediction. Publication reference: https://doi.org/10.1177/1932296819868727 |
Participant flow
Participants by arm
| Arm | Count |
|---|---|
| CGM Users User of CGM and CSII | 472 |
| Total | 472 |
Baseline characteristics
| Characteristic | CGM Users | — |
|---|---|---|
| Age, Continuous | 44 years STANDARD_DEVIATION 15 | — |
| BMI | 26 kg/m^2 STANDARD_DEVIATION 4 | — |
| HbA1c | 7.5 % of glycated hemoglobin STANDARD_DEVIATION 0.5 | — |
| Race and Ethnicity Not Collected | — | — Participants |
| Sex: Female, Male Female | 203 Participants | — |
| Sex: Female, Male Male | 269 Participants | — |
Adverse events
| Event type | EG000 affected / at risk |
|---|---|
| deaths Total, all-cause mortality | 0 / 0 |
| other Total, other adverse events | 0 / 0 |
| serious Total, serious adverse events | 0 / 0 |
Outcome results
Primary
Optimal Time Shift of Continuous Glucose Monitoring Measurements
Continuous glucose monitoring (CGM) measurements are delayed compared to blood glucose. The CGM signal is time-shifted -1 minute at a time and the mean absolute difference between CGM and blood glucose measurements are calculated at each step. The lowest mean absolute difference depicts the optimal time shift in minutes. The resultant mean absolute relative difference is provided as outcome. Publication reference: https://doi.org/10.1177/1932296819848721
Time frame: 16 weeks
| Arm | Measure | Value (MEAN) |
|---|---|---|
| CGM Users | Optimal Time Shift of Continuous Glucose Monitoring Measurements | 1.01 Percentage point change in MARD |
Secondary
Area Under the Receiver Operating Characteristics Curve of the Hypoglycemia Prediction
Area under the receiver operating characteristics curve (ROC-AUC) is a measure of the prediction capabilities of a prediction algorithm. Each point of the curve gives a sensitivity and a specificity of the prediction. Publication reference: https://doi.org/10.1177/1932296819868727
Time frame: 16 weeks
Population: Nine people from the original study did not have sufficient CGM, insulin, or meal data for the calculation of features.
| Arm | Measure | Value (NUMBER) |
|---|---|---|
| CGM Users | Area Under the Receiver Operating Characteristics Curve of the Hypoglycemia Prediction | 0.79 probability |