Systematic Evaluation of Continuous Glucose Monitoring Data
- Conditions
- Diabetes Mellitus
- Interventions
- Behavioral: glucose control (Substudy A)Diagnostic Test: hypoglycemia prediction (Substudy B)
- Registration Number
- NCT03545178
- Lead Sponsor
- Insel Gruppe AG, University Hospital Bern
- Brief Summary
This study retrospectively evaluates continuous glucose monitoring (CGM) and flash glucose monitoring (FGM) data and pursues two main objectives: First, the investigators analyze if glucose values are better controlled in the days directly before a consultation at our tertiary referral centre (so called "white coat adherence"). Second, the investigators use the collected CGM and FGM data to develop a hypoglycemia prediction model.
- Detailed Description
Substudy A.) Presence of white coat adherence in diabetic patients:
The investigators aim at evaluating the existence of a so called "white coat adherence" with regard to diabetes control, which means that blood-glucose is better controlled in the days immediately prior to a consultation at the diabetes clinic compared to the time-period further back. To analyse this phenomenon, the investigators use continuous glucose monitoring (CGM) and flash glucose monitoring (FGM) of diabetic patients and compare CGM-/FGM data of the last three days prior to the consultation with the CGM-/FGM data of the days 4-28 prior to the consultation, as well as the last seven days prior to the consultation with days 8-28 prior to the consultation.
Substudy B.) Retrospective data collection for the development and evaluation of a hypoglycemia prediction model:
Scope of the study is to use retrospective data for training and evaluation of a deep recurrent neural network based system for predicting the onset of hypoglycemic event at least 20 min ahead in time. The study aims to: I, assess the ability of deep learning algorithm to predict hypoglycemic events using the data collected during substudy 1. II, assess the ability of global model to be personalized using the data collected during sub-study 1. III, investigate the amount of "history" to be involved to achieve maximum performance in terms of prediction ability. IV, develop a global model, which can be easily further personalized to achieve optimum prediction performance per patient.
Recruitment & Eligibility
- Status
- COMPLETED
- Sex
- All
- Target Recruitment
- 384
- Diabetes mellitus
- CGM and/or FGM available for at least 50% of the time in last 4 weeks before consultation
- Written informed general consent for the retrospective analysis of data
- Pregnancy
Study & Design
- Study Type
- OBSERVATIONAL
- Study Design
- Not specified
- Arm && Interventions
Group Intervention Description Diabetic patients using CGM/FGM hypoglycemia prediction (Substudy B) Evaluation of glucose control and application of hypoglycemia prediction models in diabetic patients wearing CGM and/or FGM devices for at least 50% of the time during the last 4 weeks prior to the medical consultation. Diabetic patients using CGM/FGM glucose control (Substudy A) Evaluation of glucose control and application of hypoglycemia prediction models in diabetic patients wearing CGM and/or FGM devices for at least 50% of the time during the last 4 weeks prior to the medical consultation.
- Primary Outcome Measures
Name Time Method Hypoglycemia prediction (for Substudy B) 01.01.2013 - 31.07.2018; outcome assessed at study end Proportion of times a deep learning based algorithm can predict a hypoglycemic event (BG \<4.0 mmol/l) at least 20 min ahead in time?
Change of time in target glucose range day 0-3 compared to day 4-28 and day 0-7 compared to day 8-28 prior to consultation (for Substudy A) 01.01.2013 - 31.07.2018; outcome assessed at study end The time spent in the target glucose range from 3.9 to 10.0 mmol/l assessed by CGM/FGM.
- Secondary Outcome Measures
Name Time Method Change of average and standard deviation glucose day 0-3 compared to day 4-28 and day 0-7 compared to day 8-28 prior to consultation (for Substudy A) 01.01.2013 - 31.07.2018; outcome assessed at study end Average and standard deviation glucose levels based on CGM/FGM data
Change of time in hyperglycemia day 0-3 compared to day 4-28 and day 0-7 compared to day 8-28 prior to consultation (for Substudy A) 01.01.2013 - 31.07.2018; outcome assessed at study end The time with glucose levels in the significant hyperglycaemia, as based on CGM/FGM (glucose levels \> 13.9 mmol/l)
Sensor wearing time day 0-3 compared to day 4-28 and day 0-7 compared to day 8-28 prior to consultation (for Substudy A) 01.01.2013 - 31.07.2018; outcome assessed at study end Time CGM-/FGM sensor has been worn (%)
Change of mean amplitude of glucose excursion (MAGE) day 0-3 compared to day 4-28 and day 0-7 compared to day 8-28 prior to consultation (for Substudy A) 01.01.2013 - 31.07.2018; outcome assessed at study end The mean amplitude of glucose excursion assessed by CGM/FGM
Change of time above and below glucose target range day 0-3 compared to day 4-28 and day 0-7 compared to day 8-28 prior to consultation (for Substudy A) 01.01.2013 - 31.07.2018; outcome assessed at study end The time spent above and below the target glucose (3.9 to 10.0 mmol/l) assessed by CGM/FGM.
Change of coefficient of variation (CV) day 0-3 compared to day 4-28 and day 0-7 compared to day 8-28 prior to consultation (for Substudy A) 01.01.2013 - 31.07.2018; outcome assessed at study end Coefficient of variation (CV) based on CGM/FGM data
Change of time in hypoglycemia day 0-3 compared to day 4-28 and day 0-7 compared to day 8-28 prior to consultation (for Substudy A) 01.01.2013 - 31.07.2018; outcome assessed at study end The time with glucose levels \< 3.0 based on CGM/FGM data
Trial Locations
- Locations (1)
Inselspital, Bern University Hospital, University of Bern
🇨🇭Bern, BE, Switzerland