Analysis of “Performance of an Automated Insulin Delivery System in People Living With Type 2 Diabetes and Insulin Resistance: First Real-World Evidence in 26,427 Users”

Abstract

Automated insulin delivery (AID) is now standard of care for managing type 1 diabetes (T1D), yet its role in type 2 diabetes (T2D) is still emerging. While only one AID system has earned regulatory approval for T2D as of January 2025, real-world studies have demonstrated promising results. The article “Performance of an Automated Insulin Delivery system in people living with type 2 diabetes and insulin resistance: first real-world evidence in 26,427 users” presents a large-scale analysis of the MiniMed 780G system in T2D across 73 countries. By categorizing users into four phenotypically distinct cohorts, the study highlights the system’s adaptability, achieving consistent glycemic improvements across groups. Despite limitations in baseline data and clinical parameters, the findings reinforce AID as an effective and scalable therapy for insulin-requiring T2D.

According to the recently published 2025 American Diabetes Association (ADA) Standard of Care, automated insulin delivery (AID) “is rapidly becoming the standard of care for people with type 1 diabetes and should be the preferred method of insulin delivery in these individuals.” ¹ This recommendation has an A evidence rating, backed by many prospective clinical trials, large real-world data sets, and cost analyses.

However, when it comes to type 2 diabetes (T2D), as of January 2025, only one AID system has been approved by a major regulatory agency specifically for the management of T2D. ² Results from pivotal trials in T2D from other pump manufacturers are highly anticipated, but favorable real-world data analyses in this population have been published, each with hundreds of patients.^3,4

This manuscript titled “Performance of an Automated Insulin Delivery system in people living with type 2 diabetes and insulin resistance: first real-world evidence in 26,427 users” is a new retrospective, real-world analysis of people with T2D spanning 73 countries who initiated therapy with the Medtronic 780G (MM780G) system. ⁵ The authors deserve credit for gathering such a robust, large data set of users, especially notable given the fact that this specific AID system has yet to be approved for the management of T2D.

Furthermore, beyond sheer numbers, the study design was the first of its kind to focus on four patient populations, selected intentionally to represent the diversity of phenotypes within the category of T2D. ⁶ Participants were categorized into four cohorts: (1) users with total daily dose (TDD) ≥100 IU, (2) self-reported T2D users, (3) self-reported T2D users with TDD ≥100 IU, and (4) self-reported T2D users with TDD <100 IU. As an example of the diversity of T2D phenotypes, cohort C (self-reported T2D and TDD <100 IU) consisted of PWDs who might have beta-cell failure driving insulin deficiency, whereas the cohorts requiring >100 units/day might be more driven by insulin resistance and corresponding hyperinsulinemia.

Another related strength in the study design is the inclusion of a cohort of ALL patients with TDD >100 units, regardless of whether the user indicated they have T2D or not. This has two primary benefits by (1) capturing people with T2D who might not have correctly entered their diagnosis during the onboarding process and (2) including people with T1D who might have developed overlying T2D through insulin resistance and/or obesity, also known as “double diabetes.” ⁷

By analyzing the four cohorts across the T2D spectrum separately, the authors were able to demonstrate the adaptability of the MM780G by showing consistent time-in-range (TIR) >70% and time-below-range (TBR) <1% in each and every subgroup. This consistency across different patient phenotypes supports the MM780G as a robust option for T2D. One can also appreciate some subtle differences between the cohorts, namely that TIR was the highest in cohort D, which consisted of self-identifying T2D users with TDD <100 IU, presumably who are more sensitive to insulin. In the only analysis performed that compared glycemic outcomes before and after the initiation of therapy, the two cohorts with TDD >100 IU demonstrated nearly identical improvements in TIR (cohort A improved TIR by 15.9% and cohort C improved by 15.7%), larger than the cohorts that included patients with TDD <100 IU (cohort B improved TIR by 12.1% and cohort D improved by 11.9%).

Similar to real-world data from MM780G in patients with T1D, implementation of recommended optimal settings (ROS), such as the glycemic target (GT) of 100 mg/dL, and active insulin time (AIT) of 2 hours, resulted in the highest mean TIR of 78.7% of any subset. ⁸ However, the article does not make it clear why this TIR during ROS was only calculated for cohort C. Perhaps, it was due to the fact that this was the most insulin-resistant cohort.

In terms of other limitations of the study, the nature of real-world data gathering at scale through manufacturer cloud data servers leads to limitations in what and how data can be obtained. For example, key clinical data such as a subject’s weight and baseline HbA1c were not gathered. Non-pump-derived data such as diabetes type, age, and gender that are manually entered in by the patient are susceptible to input error.

Similarly, another key missing aspect to this data set is the lack of visibility into the subject’s history before initiating MM780G. For example, what percent of patients were previously being treated with multiple daily injections (MDI) vs insulin pump therapy? Were some pump users on other hybrid closed loop systems (HCL)? Also, what were the baseline glycemic metrics for these subjects before starting therapy with MM780G? Analysis 2 was the only portion of the study that looked at glycemia before and after initiation of therapy, but these data were limited to n = 1516 subjects, just a fraction of the overall data set of 26 427 users.

Conclusion

This real-world data study provides compelling evidence that the MM780G is effective for patients with T2D with varying insulin needs. All cohorts achieved consensus treatment targets with TIR >70% and TBR <1%, with even higher TIR when implementing ROS (GT of 100 mg/dL and AIT of 2 hours). The study aligns with prior research demonstrating AID efficacy in T1D populations. Comparisons with other AID systems suggest that MM780G yields competitive glycemic outcomes. In addition, users with high insulin requirements exhibited greater reliance on automated insulin delivery, suggesting that AID may compensate for inaccuracies in carbohydrate estimation.