The HEADWIND Study - Part 2
- Conditions
- Diabetes Mellitus, Type 1Diabetes
- Interventions
- Other: Controlled hypoglycaemic state while driving
- Registration Number
- NCT04569630
- Lead Sponsor
- Insel Gruppe AG, University Hospital Bern
- Brief Summary
To analyse driving behavior of individuals with type 1 diabetes in eu- and progressive hypoglycaemia while driving in a real car. Based on the driving variables provided by the car the investigators aim at establishing algorithms capable of discriminating eu- and hypoglycemic driving patterns using machine learning neural networks (deep machine learning classifiers).
- Detailed Description
Hypoglycaemia is among the most relevant acute complications of diabetes mellitus. During hypoglycaemia physical, psychomotor, executive and cognitive function significantly deteriorate. These are important prerequisites for safe driving. Accordingly, hypoglycaemia has consistently been shown to be associated with an increased risk of driving accidents and is, therefore, regarded as one of the relevant factors in traffic safety. Despite important developments in the field of diabetes technology, the problem of hypoglycaemia during driving persists. Automotive technology is highly dynamic, and fully autonomous driving might, in the end, resolve the issue of hypoglycemia-induced accidents. However, autonomous driving (level 4 or 5) is likely to be broadly available only to a substantially later time point than previously thought due to increasing concerns of safety associated with this technology. Therefore, solutions bridging the upcoming period by more rapidly and directly addressing the problem of hypoglycemia-associated traffic incidents are urgently needed.
On the supposition that driving behaviour differs significantly between euglycaemic state and hypoglycaemic state, the investigators assume that different driving patterns in hypoglycemia compared to euglycemia can be used to generate hypoglycemia detection models using machine learning neural networks (deep machine learning classifiers).
Recruitment & Eligibility
- Status
- COMPLETED
- Sex
- All
- Target Recruitment
- 22
- Informed consent as documented by signature
- Type 1 Diabetes mellitus as defined by WHO for at least 1 year or confirmed C-peptide negative (<100pmol/l with concomitant blood glucose >4 mmol/l)
- Age between 21-60 years
- HbA1c ≤ 9.0 %
- Functional insulin treatment with good knowledge of insulin self- management
- Passed driver's examination at least 3 years before study inclusion. Possession of a valid, definitive Swiss driver's license.
- Active driving in the last 6 months.
- Contraindications to the drug used to induce hypoglycaemia (insulin aspart), known hypersensitivity or allergy to the adhesive patch used to attach the glucose sensor.
- Pregnancy or intention to become pregnant during the course of the study, lactating women or lack of safe contraception
- Other clinically significant concomitant disease states as judged by the investigator
- Physical or psychological disease likely to interfere with the normal conduct of the study and interpretation of the study results as judged by the investigator
- Renal failure
- Hepatic dysfunction
- Coronary heart disease
- Other cardiovascular disease
- Epilepsy
- Drug or alcohol abuse
- Inability to follow the procedures of the study, e.g. due to language problems, psychological disorders, dementia, etc. of the participant
- Participation in another study with an investigational drug within the 30 days preceding and during the present study
- Total daily insulin dose >2 IU/kg/day
- Specific concomitant therapy washout requirements prior to and/or during study participation
- Current treatment with drugs known to interfere with metabolism or driving performance
Study & Design
- Study Type
- INTERVENTIONAL
- Study Design
- SINGLE_GROUP
- Arm && Interventions
Group Intervention Description Intervention group Controlled hypoglycaemic state while driving -
- Primary Outcome Measures
Name Time Method Accuracy of the HEADWIND-model: Diagnostic accuracy of the hypoglycemia warning system (HEADWIND) in detecting hypoglycemia (blood glucose < 3.9 and < 3.0 mmol/l) quantified as the area under the receiver operator characteristics curve (AUC ROC). 240 minutes Accuracy of the HEADWIND-model will be assessed using real car driving data recorded in progressive hypoglycemia and driving data will be analysed using applied machine learning technology for hypoglycemia detection.
- Secondary Outcome Measures
Name Time Method Incidence of Serious Adverse Events (SAEs Throughout the study, expected to be up to 12 months Serious Adverse Events will be recorded at each study visit.
Change of heart-rate variability 240 minutes Change of heart-rate variability during driving in hypoglycemia will be compared to euglycemia. Heart-rate variability will be measured with a holter-ecg and wearable devices.
Change of facial expression 240 minutes Change of facial expression during driving in hypoglycemia will be compared to euglycemia. Facial expression will be recorded by a camera.
Change of brake 240 minutes Change of brake during driving in hypoglycemia (\< 3.9mmol/L) will be compared to euglycemia (5.5 mmol/L). Driving parameters will be recorded by the study car.
Change of steer speed 240 minutes Change of steer speed during driving in hypoglycemia (\< 3.9mmol/L) will be compared to euglycemia (5.5 mmol/L). Driving parameters will be recorded by the study car.
Change of spinning 240 minutes Change of spinning during driving in hypoglycemia (\< 3.9mmol/L) will be compared to euglycemia (5.5 mmol/L). Driving parameters will be recorded by the study car.
Change of steer torque 240 minutes Change of steer torque during driving in hypoglycemia (\< 3.9mmol/L) will be compared to euglycemia (5.5 mmol/L). Driving parameters will be recorded by the study car.
Change of skin temperature 240 minutes Change of skin temperature during driving in hypoglycemia will be compared to euglycemia. Change of skin temperature will be measured with wearable devices and a thermal camera.
Change of electrodermal activity (EDA) 240 minutes Change of EDA during driving in hypoglycemia will be compared to euglycemia. EDA will be measured with wearable devices.
Change of growth hormone (GH) 240 minutes Change of GH before driving, during driving in euglycemia (5.5mmol/L), in hypoglycemia (\< 3.9mmol/L), severe hypoglycemia (\< 3mmol/L) and after hypoglycemia will be assessed.
Comparison CGM and HEADWIND-model regarding time-point of hypoglycemia detection 240 minutes Time point of hypoglycemia detection by CGM will be compared to time point of hypoglycemia detection by the HEADWIND-model.
Change of swerving 240 minutes Change of swerving during driving in hypoglycemia (\< 3.9mmol/L) will be compared to euglycemia (5.5 mmol/L). Driving parameters will be recorded by the study car.
Change of velocity 240 minutes Change of velocity during driving in hypoglycemia (\< 3.9mmol/L) will be compared to euglycemia (5.5 mmol/L). Driving parameters will be recorded by the study car.
Change of steer 240 minutes Change of steer during driving in hypoglycemia (\< 3.9mmol/L) will be compared to euglycemia (5.5 mmol/L). Driving parameters will be recorded by the study car.
Driving performance before and after hypoglycemia based on driving parameters (swerving, spinning, velocity, steer, brake, steer torque, steer speed) 240 minutes Based on significantly altered driving parameters in serious hypoglycemia (\< 3.0 mmol/L) driving performance based on swerving, spinning, velocity, steer, brake, steer torque and steer speed, before and after hypoglycemia will be assessed
Change of glucagon 240 minutes Change of glucagon before driving, during driving in euglycemia (5.5mmol/L), in hypoglycemia (\< 3.9mmol/L), severe hypoglycemia (\< 3mmol/L) and after hypoglycemia will be assessed.
Self-estimation of glucose and hypoglycemia 240 minutes Evaluation of self-estimated glucose during progressive hypoglycemia and correlation with measured blood glucose.
Time point of need-to-treat 240 minutes Time point of self-perceived need-to-treat (hypoglycemia) compared to time point of hypoglycemia detection by the HEADWIND-model and CGM.
Self-perception of hypoglycemia symptoms compared to baseline hypoglycemia awareness 240 minutes Correlation and comparison of perceived hypoglycemia symptoms on a scale from 0-6 (0 means better outcome) to baseline hypoglycemia awareness (Clarke-Score and Gold-Score, for both tests a score of higher or equal to 4 points indicates impaired awareness of hypoglycemia).
Pre-test perception of technology in general Throughout the study, expected to be up to 12 months Perception of technology in general will be assessed via questionnaire based self-reports (technology readiness index) measures on the 5-point Likert Scale ranging from "strongly disagree" to "strongly agree" with a scale ranging from -2 to 2 with higher values representing a better outcome (after inversion of negative items). The total score will be averaged across participants and used individually to support the interview responses when necessary.
Self-report of blood sugar level while driving (i.e. ecological momentary assessment) 240 minutes Comparison of perceived blood sugar level to measured blood glucose, perceived blood sugar level between drives (see outcome 21), and baseline hypoglycemia awareness (Clarke-Score and Gold-Score, for both tests a score of higher or equal to 4 points indicates impaired awareness of hypoglycemia).
Acceptance and use of the EWS Throughout the study, expected to be up to 12 months Acceptance and use of the EWS will be assessed via questionnaire based self-reports (questionnaire of technology use and acceptance) measured on the 7-point Likert scale from "strongly disagree" to "strongly agree" with a scale range from -3 to 3 with higher values representing a better outcome. The total score will be averaged per construct and across participants and used individually to support the interview responses when necessary.
Perceived working alliance with IVA Throughout the study, expected to be up to 12 months Perceived working alliance with the IVA will be assessed via questionnaire based self-reports (session alliance inventory) measured on the 6-point Likert scale from "not at all" to "completely" with a scale range from 0 to 5 and with higher values representing a better outcome. The total score will be averaged per construct and across participants and used individually to support the interview responses when necessary
Defining the glycemic level when driving performance is decreased 240 minutes Plasma-glucose level (mmol/L) when driving performance begins to be impaired will be assessed based on significantly altered driving parameters in serious hypoglycemia (\< 3.0 mmol/L) compared to euglycemia (5.5mmol/L).
Change of heart-rate 240 minutes Change of heart-rate during driving in hypoglycemia will be compared to euglycemia. Change of heart-rate will be measured with a holter-ecg and wearable devices.
Diagnostic accuracy in detecting hypoglycemia (blood glucose <3.9 mmol/l and <3.0 mmol/l) and hyperglycemia (blood glucose >13.9 mmol/l and >16.7 mmol/l) quantified as the area under the receiver operator characteristics curve using physiological data Throughout the study, expected to be up to 12 months Accuracy of dysglycemia detection using physiological data (heart-rate, heart-rate variability, skin temperature, EDA) recorded with wearable devices during the study period will be analysed using applied machine learning technology.
CGM time-delay during the controlled hypoglycemic state 240 minutes Time-delay (minutes) of CGM Sensor (Dexcom G6) during progressive hypoglycemia (hypoglycemic clamp) will be assessed compared to plasma glucose.
Change of catecholamines 240 minutes Change of catecholamines before driving, during driving in euglycemia (5.5mmol/L), in hypoglycemia (\< 3.9mmol/L), severe hypoglycemia (\< 3mmol/L) and after hypoglycemia will be assessed.
Accuracy-comparison of HEADWIND-model and HEADWINDplus-model 240 minutes Diagnostic accuracy of the hypoglycaemia warning system (HEADWIND) to detect hypoglycaemia (blood glucose \< 3.9 mmol/l and \< 3.0 mmol/l) quantified as the area under the receiver operator characteristics curve (AUC ROC) using only driving parameters (HEADWIND-model) will be compared to the HEADWIND-model with the additional integration of physiological parameters, video and eye tracker data, in particular heart-rate, heart-rate variability, electrodermal activity (EDA), skin temperature and facial expression (HEADWINDplus-model)
Self-estimation of driving performance 240 minutes Evaluation of self-estimated driving-performance in severe hypoglycemia (\< 3.0 mmol/L) compared to euglycemia (5.5mmol/L). Self-estimated driving performance will be assessed on a absolute 7-point scale from 0-6 (a lower value means better outcome).
Driving mishaps and interventions by the driving instructor in euglycaemia (5-8 mmol/l), hypoglycaemia (< 3.9 mmol/l) and severe hypoglycaemia (< 3.0 mmol/l). 240 minutes Driving mishaps and interventions will be assessed by the driving instructor using an assessment questionnaire with 4 questions on a 7 point Likert scale (lower value means worse outcome)
Change of eye movement 240 minutes Change of eye movement and gaze behaviour during driving in hypoglycemia will be compared to euglycemia. Eye movement of the participant will be recorded by a camera and an eye-tracker.
Comparison CGM and HEADWIND-model regarding glycemia 240 minutes Blood glucose at time point of hypoglycemia detection by the HEADWIND- model compared to glucose value of CGM at same time point will be assessed.
Direct comparison between IVA's prompts and the behavioral responses 240 minutes Direct comparison of conversational turns between IVA and patient during the ecological momentary assessment and the hypoglycaemia support.
Diagnostic accuracy in detecting hypoglycemia (blood glucose < 3.9 mmol/l and < 3.0 mmol/l) quantified as the area under the receiver operator curve (AUC-ROC) using video data Throughout the study, expected to be up to 12 months Using video data recorded by a camera and a thermal camera accuracy in hypoglycaemia detection will be analysed with applied machine learning technology.
Diagnostic accuracy in detecting hypoglycemia (blood glucose < 3.9 mmol/l and < 3.0 mmol/l) quantified as the area under the receiver operator curve (AUC-ROC) using eye-tracking data Throughout the study, expected to be up to 12 months Using eye-tracking data recorded by a camera and an eye-tracker (to record gaze behaviour) accuracy in hypoglycemia detection will be analysed with applied machine learning technology.
Change of cortisol 240 minutes Change of cortisol before driving, during driving in euglycemia (5.5mmol/L), in hypoglycemia (\< 3.9mmol/L), severe hypoglycemia (\< 3mmol/L) and after hypoglycemia will be assessed.
Change of insulin 240 minutes Insulin levels will be measured before driving, during driving in euglycemia (5.5mmol/L), in hypoglycemia (\< 3.9mmol/L), serious hypoglycemia (\< 3mmol/L) and after hypoglycemia will be assessed.
Glycemic level at time point of hypoglycemia detection by the HEADWIND-model 240 minutes Blood glucose at time point of hypoglycemia detection by the HEADWIND-model will be determined.
Self-perception of hypoglycemia symptoms 240 minutes Correlation of perceived hypoglycemia symptoms on a scale from 0-6 (0 means better outcome) to measured blood glucose.
Incidence of Adverse Events (AEs) Throughout the study, expected to be up to 12 months Adverse Events will be recorded at each study visit.
CGM accuracy during the controlled hypoglycemic state 240 minutes Accuracy (mean absolute relative difference, MARD) of CGM Sensor (Dexcom G6) in euglycemia (3.9 - 10 mmol/L), hypoglycemia (3.0 - 3.9mmol/L) and severe hypoglycemia (\< 3.0 mmol/L) will be assessed based on plasma glucose measurements
Pre-test experience with in-vehicle voice assistants (IVAs) and technology in general Throughout the study, expected to be up to 12 months Pre-test experience with IVAs and technology in general will be assessed via questionnaire based self-reports (questionnaire of technology use and acceptance). The constructs Performance expectancy, Effort expectancy, Social influence, Facilitating conditions, Hedonic motivation, and Behavioural intention are measured on the 7-point Likert scale from "strongly disagree" to "strongly agree" with a scale range from -3 to 3 with higher values representing a better outcome. The construct Use is measured on the 7-point Likert scale ranging from "never" to "always" with a scale range from -3 to 3. The total score will be averaged per construct and across participants and used individually to support the interview responses when necessary.
Direct comparison of driving performance scores assessed by the driving instructor in euglycemia (5-8 mmol/l), hypoglycaemia (<3.9 mmol/l) and severe hypoglycaemia (< 3.0 mmol/l) 240 minutes Driving performance will be assessed by the driving instructor using an assessment questionnaire with a score from 1 to 7 (7 means the best outcome)
Comparison of cognitive trust in competence and session alliance with IVA to warning type 240 minutes Cognitive trust in competence with IVA will be assessed via questionnaire based self-reports (Cognitive trust in competence construct from Trust and adoption of recommendations agents questionnaire), measured on the 7-point Likert scale from "strongly disagree" to "strongly agree" with a scale range from -3 to 3 and with higher values representing a better outcome. Session alliance with IVA will be assessed via questionnaire based self-reports (item from Session Alliance Inventory), measured on the 6-point Likert scale from "not at all" to "completely" with a scale range from 0 to 5 and with higher values representing a better outcome. The questionnaire will be submitted after delivering IVA's support intervention and will be compared with the type of warning delivered (i.e. disclosure vs no disclosure).
General user experience of the early hypoglycaemia warning system (EWS) Throughout the study, expected to be up to 12 months General user experience of the EWS will be assessed via questionnaire based self-reports (questionnaire for User experience questionnaire and van der Laan scale) measured on an analogue scale with adjective at its extremes (e.g. easy to learn-hard to learn, boring-exciting, good-bad, etc.) with a scale range from 0 to 100. The scores will be averaged for each scale across participants and used individually to support the interview responses when necessary.
Cognitive trust in competence and emotional trust in the recommendations from IVA Throughout the study, expected to be up to 12 months Cognitive trust in competence and emotional trust in the recommendations from IVA will be assessed via questionnaire based self-reports (Cognitive trust in competence and emotional trust constructs from Trust and adoption of recommendations agents questionnaire) measured on the 7-point Likert scale from "strongly disagree" to "strongly agree" with a scale range from -3 to 3 and with higher values representing a better outcome. The total score will be averaged per construct and across participants and used individually to support the interview responses when necessary
Trial Locations
- Locations (1)
Department of Diabetes, Endocrinology, Nutritional Medicine and Metabolism
🇨🇭Bern, Switzerland