Type 1 Diabetes
Conditions
Keywords
automated insulin delivery systems, Context patterns
Brief summary
Automated Insulin Delivery (AID) systems have now become an important standard-of-care for people with T1D and have demonstrated a reduction, but not elimination, of hypoglycemia during long-term studies. One limitation of current AID systems is that they have no knowledge about the context or environment that a person is currently experiencing. Contextual patterns can potentially improve the performance of an AID by recognizing environments or patterns of living that are related to changes in glucose. The team at OHSU is developing a context-aware glucose prediction algorithm that will capture context data from the patient both indoors and outdoors. This context data will be provided to the algorithm to allow for detecting contextual patterns that might relate to high or low glucose. The goal of this study will be the creation of a data set that will include contextual patterns along with glucose, insulin and physiological data.
Detailed description
Subjects will be on study for 28 days. Sensor glucose, activity, exercise, insulin, indoor and outdoor contextual patterns and meal data will be collected during this time. Subjects will wear the Dexcom G6 CGM system and a physical activity monitor for the entire 28 days. Subjects will continue to use their own insulin pump. Subjects will be asked to also wear a MotioWear indoor/outdoor context-aware tracking tag and to install the MotioWear beacons within their home. Subjects will be randomized to complete either aerobic, high intensity interval training, or resistance exercise videos twice weekly at home during weeks 1 and 2 and once during weeks 3 and 4. Subjects will also ingest a self-selected meal prior to these prescribed exercise sessions. Subjects will eat a high carbohydrate dinner once each week on the same day at the same approximate time of day (but not on the exercise days). Subjects will use the T1 DEXI mobile app created by OHSU to capture meal and exercise data along with photos of meals the day of exercise and the day after. While at home, subjects will check CBG before and after exercise, for symptoms of hypoglycemia, and for Dexcom G6 alarms for sensor \<70 mg/dL and \>250 mg/dL.
Interventions
OTHERExercise
Subjects will be randomized to complete either aerobic, high intensity interval training, or resistance exercise videos twice weekly at home.
Sponsors
National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK)
Oregon Health and Science University
Study design
Allocation
RANDOMIZED
Intervention model
PARALLEL
Primary purpose
BASIC_SCIENCE
Masking
NONE
Eligibility
Sex/Gender
ALL
Age
18 Years to 65 Years
Healthy volunteers
No
Inclusion criteria
* Diagnosis of type 1 diabetes mellitus for at least 1 year. * Male or female subjects 18 to 65 years of age. * Physically willing and able to perform 30 min of exercise (as determined by the investigator after reviewing the subject's activity level). * Current use of an insulin pump for at least 3 months. * A1C \<10.5% at the time of screening. * Willingness to follow all study procedures, including attending all clinic visits. * Willingness to sign informed consent and HIPAA documents.
Exclusion criteria
* Female of childbearing potential who is pregnant or intending to become pregnant or breast-feeding, or is not using adequate contraceptive methods. Acceptable contraception includes birth control pill / patch / vaginal ring, Depo-Provera, Norplant, an IUD, the double barrier method (the woman uses a diaphragm and spermicide and the man uses a condom), or abstinence. * Any cardiovascular disease, defined as a clinically significant EKG abnormality at the time of screening or any history of: stroke, heart failure, myocardial infarction, angina pectoris, or coronary arterial bypass graft or angioplasty. Diagnosis of 2nd or 3rd degree heart block or any non-physiological arrhythmia judged by the investigator to be exclusionary. * Renal insufficiency (GFR \< 60 ml/min, using the MDRD equation as reported by the OHSU laboratory). * Liver failure, cirrhosis, or any other liver disease that compromises liver function as determined by the investigator. * Hematocrit of less than 36% for men, less than 32% for women. * History of severe hypoglycemia during the past 12 months prior to screening visit or hypoglycemia unawareness as judged by the investigator. Subjects will complete a hypoglycemia awareness questionnaire. Subjects will be excluded for four or more R responses. * Adrenal insufficiency. * Any active infection. * Known or suspected abuse of alcohol, narcotics, or illicit drugs. * Seizure disorder. * Active foot ulceration. * Peripheral arterial disease. * Major surgical operation within 30 days prior to screening. * Use of an investigational drug within 30 days prior to screening. * Chronic usage of any immunosuppressive medication (such as cyclosporine, azathioprine, sirolimus, or tacrolimus). * Bleeding disorder or platelet count below 50,000. * Current administration of oral or parenteral corticosteroids. * Any life threatening disease, including malignant neoplasms and medical history of malignant neoplasms within the past 5 years prior to screening (except basal and squamous cell skin cancer). * Beta blockers or non-dihydropyridine calcium channel blockers. * Current use of any medication intended to lower glucose other than insulin (ex. use of liraglutide). * A positive response to any of the questions from the Physical Activity Readiness Questionnaire with one exception: subject will not be excluded if he/she takes a single blood pressure medication that doesn't impact heart rate and blood pressure is controlled on the medication (blood pressure is less than 140/90 mmHg). * Any chest discomfort with physical activity, including pain or pressure, or other types of discomfort. * Any clinically significant disease or disorder which in the opinion of the Investigator may jeopardize the subject's safety or compliance with the protocol.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Comparison of the Mean Absolute Relative Difference When Including Patterns in Hypoglycemia Prediction. | 28 days | We used our recently published long short term memory neural network (LSTM) to predict sensor glucose 30-minutes in advance across the entire 4-week study duration when glucose was \< 70 mg/dL. We then used the context-aware pattern recognition algorithm to predict when hypoglycemia would occur 30-minutes in the future, and if hypoglycemia was predicted, we included a bias correction that is specific to the hypogylcemia region of glucose measurements. The outcome measure shows the reduction of MARD when the LSTM is corrected using the pattern-based bias correction algorithm. MARD is calculated by subtracting the new sensor glucose - reference value dividing by the reference value. A negative value means that the MARD was reduced. The LSTM is being compared with Dexcom G6 CGM values to determine the MARD. Physiologically relevant thresholds are less than 55 mg/dl, less than 70 mg/dl, above 180 mg/dl and above 250 mg/dl. The Dexcom G6 target range is 70-180 mg/dl. |
| Comparison of the Mean Relative Difference When Including Patterns in Hypoglycemia Prediction. | 28 days | We used our recently published long short term memory neural network (LSTM) to predict glucose 30-minutes in advance across the entire 4-week study duration when glucose was \< 70 mg/dL. We then used the context-aware pattern recognition algorithm to predict when hypoglycemia would occur 30-minutes in the future, and if hypoglycemia was predicted, we included a bias correction that is specific to the hypogylcemia region of glucose measurements. The outcome measure shows the reduction of mean relative difference (MRD) when the LSTM is corrected using the pattern-based bias correction algorithm. MARD is calculated by subtracting the new sensor glucose - reference value dividing by the reference value. A negative value means that the MARD was reduced. The LSTM is being compared with Dexcom G6 CGM values to determine the MARD. Physiologically relevant thresholds are less than 55 mg/dl, less than 70 mg/dl, above 180 mg/dl and above 250 mg/dl. The Dexcom G6 target range is 70-180 mg/dl. |
Countries
United States
Participant flow
Participants by arm
| Arm | Count |
|---|---|
| Aerobic Exercise Exercise: Subjects will be randomized to complete either aerobic, high intensity interval training, or resistance exercise videos twice weekly at home. | 10 |
| Resistance Exercise Exercise: Subjects will be randomized to complete either aerobic, high intensity interval training, or resistance exercise videos twice weekly at home. | 10 |
| High Intensity Interval Exercise Exercise: Subjects will be randomized to complete either aerobic, high intensity interval training, or resistance exercise videos twice weekly at home. | 10 |
| Total | 30 |
Withdrawals & dropouts
| Period | Reason | FG000 | FG001 | FG002 |
|---|---|---|---|---|
| At Home Period (28 Days) | Withdrawal by Subject | 0 | 0 | 1 |
Baseline characteristics
| Characteristic | Aerobic Exercise | Resistance Exercise | High Intensity Interval Exercise | Total |
|---|---|---|---|---|
| Age, Categorical <=18 years | 0 Participants | 0 Participants | 0 Participants | 0 Participants |
| Age, Categorical >=65 years | 0 Participants | 0 Participants | 0 Participants | 0 Participants |
| Age, Categorical Between 18 and 65 years | 10 Participants | 10 Participants | 10 Participants | 30 Participants |
| Age, Continuous | 37.1 years STANDARD_DEVIATION 12.9 | 36.5 years STANDARD_DEVIATION 12.9 | 34.7 years STANDARD_DEVIATION 12.3 | 36.1 years STANDARD_DEVIATION 12.3 |
| Ethnicity (NIH/OMB) Hispanic or Latino | 1 Participants | 0 Participants | 0 Participants | 1 Participants |
| Ethnicity (NIH/OMB) Not Hispanic or Latino | 9 Participants | 10 Participants | 10 Participants | 29 Participants |
| Ethnicity (NIH/OMB) Unknown or Not Reported | 0 Participants | 0 Participants | 0 Participants | 0 Participants |
| Race (NIH/OMB) American Indian or Alaska Native | 0 Participants | 1 Participants | 0 Participants | 1 Participants |
| Race (NIH/OMB) Asian | 1 Participants | 1 Participants | 1 Participants | 3 Participants |
| Race (NIH/OMB) Black or African American | 0 Participants | 0 Participants | 0 Participants | 0 Participants |
| Race (NIH/OMB) More than one race | 0 Participants | 0 Participants | 1 Participants | 1 Participants |
| Race (NIH/OMB) Native Hawaiian or Other Pacific Islander | 0 Participants | 0 Participants | 0 Participants | 0 Participants |
| Race (NIH/OMB) Unknown or Not Reported | 0 Participants | 0 Participants | 0 Participants | 0 Participants |
| Race (NIH/OMB) White | 9 Participants | 8 Participants | 8 Participants | 25 Participants |
| Region of Enrollment United States | 10 participants | 10 participants | 10 participants | 30 participants |
| Sex: Female, Male Female | 7 Participants | 6 Participants | 5 Participants | 18 Participants |
| Sex: Female, Male Male | 3 Participants | 4 Participants | 5 Participants | 12 Participants |
Adverse events
| Event type | EG000 affected / at risk | EG001 affected / at risk | EG002 affected / at risk |
|---|---|---|---|
| deaths Total, all-cause mortality | 0 / 10 | 0 / 10 | 0 / 10 |
| other Total, other adverse events | 3 / 10 | 1 / 10 | 3 / 10 |
| serious Total, serious adverse events | 0 / 10 | 0 / 10 | 0 / 10 |
Outcome results
Primary
Comparison of the Mean Absolute Relative Difference When Including Patterns in Hypoglycemia Prediction.
We used our recently published long short term memory neural network (LSTM) to predict sensor glucose 30-minutes in advance across the entire 4-week study duration when glucose was \< 70 mg/dL. We then used the context-aware pattern recognition algorithm to predict when hypoglycemia would occur 30-minutes in the future, and if hypoglycemia was predicted, we included a bias correction that is specific to the hypogylcemia region of glucose measurements. The outcome measure shows the reduction of MARD when the LSTM is corrected using the pattern-based bias correction algorithm. MARD is calculated by subtracting the new sensor glucose - reference value dividing by the reference value. A negative value means that the MARD was reduced. The LSTM is being compared with Dexcom G6 CGM values to determine the MARD. Physiologically relevant thresholds are less than 55 mg/dl, less than 70 mg/dl, above 180 mg/dl and above 250 mg/dl. The Dexcom G6 target range is 70-180 mg/dl.
Time frame: 28 days
Population: Each arm had participants that completed the study but had incomplete insulin data which prevented inclusion in the analysis of outcome measures: 5 from aerobic arm, 4 from resistance arm, 3 from HIIT arm.
| Arm | Measure | Value (MEAN) | Dispersion |
|---|---|---|---|
| Aerobic Exercise | Comparison of the Mean Absolute Relative Difference When Including Patterns in Hypoglycemia Prediction. | -0.5 percent difference of MARD | Standard Deviation 7.6 |
| Resistance Exercise | Comparison of the Mean Absolute Relative Difference When Including Patterns in Hypoglycemia Prediction. | -3.1 percent difference of MARD | Standard Deviation 1.4 |
| High Intensity Interval (HIIT) Exercise | Comparison of the Mean Absolute Relative Difference When Including Patterns in Hypoglycemia Prediction. | -2.8 percent difference of MARD | Standard Deviation 2.4 |
Primary
Comparison of the Mean Relative Difference When Including Patterns in Hypoglycemia Prediction.
We used our recently published long short term memory neural network (LSTM) to predict glucose 30-minutes in advance across the entire 4-week study duration when glucose was \< 70 mg/dL. We then used the context-aware pattern recognition algorithm to predict when hypoglycemia would occur 30-minutes in the future, and if hypoglycemia was predicted, we included a bias correction that is specific to the hypogylcemia region of glucose measurements. The outcome measure shows the reduction of mean relative difference (MRD) when the LSTM is corrected using the pattern-based bias correction algorithm. MARD is calculated by subtracting the new sensor glucose - reference value dividing by the reference value. A negative value means that the MARD was reduced. The LSTM is being compared with Dexcom G6 CGM values to determine the MARD. Physiologically relevant thresholds are less than 55 mg/dl, less than 70 mg/dl, above 180 mg/dl and above 250 mg/dl. The Dexcom G6 target range is 70-180 mg/dl.
Time frame: 28 days
| Arm | Measure | Value (MEAN) | Dispersion |
|---|---|---|---|
| Aerobic Exercise | Comparison of the Mean Relative Difference When Including Patterns in Hypoglycemia Prediction. | -8.1 percent difference of MRD | Standard Deviation 5.1 |
| Resistance Exercise | Comparison of the Mean Relative Difference When Including Patterns in Hypoglycemia Prediction. | -4.8 percent difference of MRD | Standard Deviation 2.4 |
| High Intensity Interval (HIIT) Exercise | Comparison of the Mean Relative Difference When Including Patterns in Hypoglycemia Prediction. | -4.0 percent difference of MRD | Standard Deviation 3.2 |