Skip to content
type1.science
N/A (device / behavioral)CompletedACTRN12620000034932

CREATE: open-source automated insulin delivery vs sensor-augmented pump

What this study tests

A randomized trial of a specific open-source AID configuration: OpenAPS in modified AndroidAPS with a preproduction DANA-i pump and Dexcom G6. In 97 children and adults with T1D, time in range improved versus sensor-augmented pump therapy over 24 weeks. Neither group had severe hypoglycemia or DKA. This trial does not establish safety for every open-source system.

Editorial review: .

Most recent recorded citation date: 2023-02-28. Only explicit date metadata is included; an undated citation may be newer. This does not mean every claim was reviewed on that date.

Trial status, labels and access can change between reviews. How we review the evidence · How to read the evidence

Full sources ↓Full discussion ↓Report an issue →

Evidence at a glance

Who can enter the study?
Children aged 7–15 and adults aged 16–70 with type 1 diabetes, recruited in New Zealand. See the registered protocol for complete eligibility requirements. Registered eligibility ↗Enrollment criteria describe who may enter. They do not establish who was analyzed or the denominator for a reported result.
Reported benefit and results
At 24 weeks, time in range rose from 61.2% to 71.2% with AID and fell from 57.7% to 54.5% with sensor-augmented pump therapy: adjusted difference 14 percentage points (95% CI 9.2 to 18.8; P<0.001). The paper reports 3 hours 21 minutes more daily time in range. No treatment-effect difference by age was detected (P=0.56); neither arm had severe hypoglycemia or DKA. Two AID participants withdrew because of connectivity issues. In the 24-week continuation, controls crossed to AID; a model combining between-group and within-person comparisons estimated a 12.2-point treatment effect (95% CI 10.4 to 14.1). This is not a further 48-week randomized comparison.Read the result sources and their limitations →
Important harms
Read the reported results, safety discussion and original sources. A separate harms summary is not available; this does not establish safety.
Research access and approval
Study regions: Other regions. A trial listing does not establish product approval or current recruitment at a particular site. Check the study’s current entry requirements.
What remains uncertain?
Read the full discussion and original sources for study limitations. Planned endpoints and completion dates are not results.

Research status alone does not establish approval, clinical benefit or local availability.

Primary endpoints

  • Percentage of time in target glucose range (70-180 mg/dL; 3.9-10.0 mmol/L) during the final 2 weeks (days 155-168) of the 24-week trial

The full picture

Trial and results

CREATE randomized 97 participants (48 children and 49 adults): 44 to open-source AID and 53 to sensor-augmented pump therapy. The tested system combined modified AndroidAPS 2.8, OpenAPS 0.7.0, a preproduction DANA-i pump and Dexcom G6. The primary outcome was time in range (70–180 mg/dL) during days 155–168.2

At 24 weeks, time in range increased from 61.2% to 71.2% in the AID arm and decreased from 57.7% to 54.5% in controls. The adjusted advantage was 14 percentage points (95% CI 9.2 to 18.8; P<0.001). Neither group had severe hypoglycemia or DKA; two AID participants withdrew because of connectivity issues.2

Continuation study

Both arms entered a further 24-week phase using AID; the original controls crossed over. Most continuation participants (69/94) used a preproduction YpsoPump. The analysis combined between-group and within-person comparisons across the complete dataset, estimating a 12.2-point TIR treatment effect (95% CI 10.4 to 14.1). It is not a 48-week parallel-group randomized effect.4

No severe hypoglycemia or DKA occurred with either treatment over 48 weeks. Six participants from the original control arm withdrew during continuation: three for hardware issues, two preferring sensor-augmented pump therapy, and one for infusion-site irritation. These results support this study configuration; they cannot establish rare-event safety or equivalent performance for every open-source build.4

Sources

  1. [1]
  2. [2]
    Open-Source Automated Insulin Delivery in Type 1 Diabetes. · Peer-reviewed study · 2022-09-01

    Burnside MJ, et al. Open-Source Automated Insulin Delivery in Type 1 Diabetes. New England Journal of Medicine (2022);387:869–881.

  3. [3]
  4. [4]
    Extended Use of an Open-Source Automated Insulin Delivery System in Children and Adults with Type 1 Diabetes: The 24-Week Continuation Phase Following the CREATE Randomized Controlled Trial. · Peer-reviewed study · 2023-02-28

    Burnside MJ, et al. Extended Use of an Open-Source Automated Insulin Delivery System in Children and Adults with Type 1 Diabetes: The 24-Week Continuation Phase Following the CREATE Randomized Controlled Trial. Diabetes Technology & Therapeutics (2023);25:250–259.

  5. [5]