Trial Outcomes & Findings for Development of Algorithms for a Hypoglycemic Prevention Alarm: Closed Loop Study (NCT NCT00884611)

NCT ID: NCT00884611

Last Updated: 2018-02-28

Results Overview

Nights with CGM sensor values \< 60 mg/dL were considered to be undesirable. A Kalman filter-based model algorithm predicted whether the sensor glucose level would fall below 80 mg/dL and would suspend insulin delivery as needed. Participants may have received treatment using one or more of the following algorithms: Algorithm 1 had a hypoglycaemic prediction horizon of 70 minutes; algorithm 2: 50 minutes; algorithm 3: 30 minutes.

• Recruitment status: COMPLETED
• Study phase: NA
• Target enrollment: 20 participants
• Primary outcome timeframe: 21 days
• Results posted on: 2018-02-28

Participant Flow

Participant milestones

Measure	Predictive Suspend Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.
Overall Study STARTED	20
Overall Study Enrolled, Not Treated	1
Overall Study COMPLETED	19
Overall Study NOT COMPLETED	1

Reasons for withdrawal

Withdrawal data not reported

Baseline Characteristics

Development of Algorithms for a Hypoglycemic Prevention Alarm: Closed Loop Study

Baseline characteristics by cohort

Measure	Predictive Suspend n=20 Participants Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.
Age, Categorical <=18 years	0 Participants n=5 Participants
Age, Categorical Between 18 and 65 years	20 Participants n=5 Participants
Age, Categorical >=65 years	0 Participants n=5 Participants
Sex: Female, Male Female	9 Participants n=5 Participants
Sex: Female, Male Male	11 Participants n=5 Participants

PRIMARY outcome

Timeframe: 21 days

Population: Participants who were treated and had data for the respective algorithm were included in the analysis.

Nights with CGM sensor values < 60 mg/dL were considered to be undesirable. A Kalman filter-based model algorithm predicted whether the sensor glucose level would fall below 80 mg/dL and would suspend insulin delivery as needed. Participants may have received treatment using one or more of the following algorithms: Algorithm 1 had a hypoglycaemic prediction horizon of 70 minutes; algorithm 2: 50 minutes; algorithm 3: 30 minutes.

Outcome measures

Measure	Algorithm 1 - Control Nights n=38 Nights Algorithm 1 had a hypoglycaemic prediction horizon of 70 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 1 - Intervention Nights n=67 Nights Algorithm 1 had a hypoglycaemic prediction horizon of 70 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 2 - Control Nights n=48 Nights Algorithm 2 had a hypoglycaemic prediction horizon of 50 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 2 - Intervention Nights n=108 Nights Algorithm 2 had a hypoglycaemic prediction horizon of 50 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 3 - Control Nights n=37 Nights Algorithm 3 had a hypoglycaemic prediction horizon of 30 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 3 - Intervention Nights n=77 Nights Algorithm 3 had a hypoglycaemic prediction horizon of 30 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.
Percentage of Nights With CGM (Continuous Glucose Monitor) Sensor Values < 60 mg/dL	24 percentage of nights	12 percentage of nights	25 percentage of nights	11 percentage of nights	22 percentage of nights	8 percentage of nights

SECONDARY outcome

Timeframe: 21 days

Population: Participants who were treated and had data for the respective algorithm were included in the analysis.

Nights with CGM sensor values >180 mg/dL were considered to be undesirable. Participants may have received treatment using one or more of the following algorithms: Algorithm 1 had a hypoglycaemic prediction horizon of 70 minutes; algorithm 2: 50 minutes; algorithm 3: 30 minutes.

Outcome measures

Measure	Algorithm 1 - Control Nights n=38 Nights Algorithm 1 had a hypoglycaemic prediction horizon of 70 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 1 - Intervention Nights n=67 Nights Algorithm 1 had a hypoglycaemic prediction horizon of 70 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 2 - Control Nights n=48 Nights Algorithm 2 had a hypoglycaemic prediction horizon of 50 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 2 - Intervention Nights n=108 Nights Algorithm 2 had a hypoglycaemic prediction horizon of 50 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 3 - Control Nights n=37 Nights Algorithm 3 had a hypoglycaemic prediction horizon of 30 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 3 - Intervention Nights n=77 Nights Algorithm 3 had a hypoglycaemic prediction horizon of 30 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.
Percentage of Nights With CGM Values >180 mg/dL	63 percentage of nights	78 percentage of nights	29 percentage of nights	56 percentage of nights	49 percentage of nights	60 percentage of nights

SECONDARY outcome

Timeframe: 21 days

Population: Participants who were treated and had data for the respective algorithm were included in the analysis.

Desirable glucose level was 70-180 mg/mL. Average of all morning BG data is presented. Participants may have received treatment using one or more of the following algorithms: Algorithm 1 had a hypoglycaemic prediction horizon of 70 minutes; algorithm 2: 50 minutes; algorithm 3: 30 minutes.

Outcome measures

Measure	Algorithm 1 - Control Nights n=5 Participants Algorithm 1 had a hypoglycaemic prediction horizon of 70 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 1 - Intervention Nights n=5 Participants Algorithm 1 had a hypoglycaemic prediction horizon of 70 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 2 - Control Nights n=12 Participants Algorithm 2 had a hypoglycaemic prediction horizon of 50 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 2 - Intervention Nights n=12 Participants Algorithm 2 had a hypoglycaemic prediction horizon of 50 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 3 - Control Nights n=9 Participants Algorithm 3 had a hypoglycaemic prediction horizon of 30 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.	Algorithm 3 - Intervention Nights n=9 Participants Algorithm 3 had a hypoglycaemic prediction horizon of 30 minutes. Participants had continuous glucose monitoring (CGM) using a glucose sensor and received insulin from an insulin pump during sleep. On intervention nights, participants received insulin uising an algorithm that allowed a computer to assess the data received from the CGM sensor and suspend insulin delivery to avoid potential hypoglycaemia. On control night, participants received insulin delivery as normally received by the insulin pump. Participants had 2 intervention nights to each control night.
Mean Morning Blood Glucose (BG)	125 mg/dL Standard Deviation 53	158 mg/dL Standard Deviation 52	138 mg/dL Standard Deviation 63	151 mg/dL Standard Deviation 57	133 mg/dL Standard Deviation 57	144 mg/dL Standard Deviation 48

Adverse Events

Control Night (PLGS Algorithm OFF)

Serious events: 0 serious events

Other events: 0 other events

Deaths: 0 deaths

Intervention Night (PLGS Algorithm ON)

Serious events: 0 serious events

Other events: 0 other events

Deaths: 0 deaths

Serious adverse events

Adverse event data not reported

Other adverse events

Adverse event data not reported

Additional Information

Dr. Bruce Buckingham, MD

Stanford University

Email: buckingham@stanford.edu

Results disclosure agreements

• Principal investigator is a sponsor employee
• Publication restrictions are in place