Closed-loop insulin delivery: understanding when and how it is effective

Performance and patients' satisfaction with the A7+TouchCare insulin patch pump system: A randomized controlled non-inferiority study

BACKGROUND

We assessed the performance and patient satisfaction of a new insulin patch pump, the A7+TouchCare (Medtrum), compared with the Omnipod system.

METHODS

This multicenter, randomized, open-label, controlled study enrolled 100 adult patients with type 1 or type 2 diabetes mellitus (A1C ≥ 6.5% and ≤ 9.5%, i.e., 48 to 80 mmol/mol) who were assigned with the Omnipod or with the A7+TouchCare pump for 3 months. The primary study outcome was the glucose management indicator (GMI) calculated with continuous glucose monitoring (CGM).

RESULTS

Premature withdrawals occurs respectively in 2 and 9 participants in the Omnipod and TouchCare groups. In the Per Protocol analysis, the difference in GMI between groups was 0.002% (95% confidence interval -0.251; 0.255). The non-inferiority was demonstrated since the difference between treatments did not overlap the pre-defined non-inferiority margin (0.4%). There was no significant difference in CGM parameters between groups. On average, patients in both groups were satisfied/very satisfied with the insulin pump system. Patients preferred Omnipod as an insulin management system and especially the patch delivery system but preferred the A7+TouchCare personal diabetes manager to control the system.

CONCLUSIONS

This study showed that the A7+TouchCare insulin pump was as efficient as the Omnipod pump in terms of performance and satisfaction.

CLINICAL TRAIL REGISTRATION

The study was registered in the ClinicalTrials.gov protocol register (NCT04223973).

Long-Term Glucose Forecasting for Open-Source Automated Insulin Delivery Systems: A Machine Learning Study with Real-World Variability Analysis

Glucose forecasting serves as a backbone for several healthcare applications, including real-time insulin dosing in people with diabetes and physical activity optimization. This paper presents a study on the use of machine learning (ML) and deep learning (DL) methods for predicting glucose variability (GV) in individuals with open-source automated insulin delivery systems (AID). A three-stage experimental framework is employed in this work to systematically implement and evaluate ML/DL methods on a large-scale diabetes dataset collected from individuals with open-source AID. The first stage involves data collection, the second stage involves data preparation and exploratory analysis, and the third stage involves developing, fine-tuning, and evaluating ML/DL models. The performance and resource costs of the models are evaluated alongside relative and proportional errors for 17 GV metrics. Evaluation of fine-tuned ML/DL models shows considerable accuracy in glucose forecasting and variability analysis up to 48 h in advance. The average MAE ranges from 2.50 mg/dL for long short-term memory models (LSTM) to 4.94 mg/dL for autoregressive integrated moving average (ARIMA) models, and the RMSE ranges from 3.7 mg/dL for LSTM to 7.67 mg/dL for ARIMA. Model execution time is proportional to the amount of data used for training, with long short-term memory models having the lowest execution time but the highest memory consumption compared to other models. This work successfully incorporates the use of appropriate programming frameworks, concurrency-enhancing tools, and resource and storage cost estimators to encourage the sustainable use of ML/DL in real-world AID systems.

Comparing Outcomes and Cost Between Pancreas Transplant and Standard of Care in Patients With Type 1 Diabetes

OBJECTIVES

Pancreas transplant is the only treatment that establishes normal glucose levels for patients diagnosed with diabetes. We analyzed the outcome of pancreas transplant alone (PTA) versus standard of care in the United States from 2008 to 2018. We also developed an economic model to analyze the cost-effectiveness of pancreas transplant versus continuing standard of care.

METHODS

We used the Scientific Registry of Transplant Recipients database and analyzed PTA recipient survival. Using those results, we developed a Markov model that followed a cohort of 40-year-old patients with type 1 diabetes over a 10-year time horizon. The primary outcomes were (i) the survival benefit of a pancreas transplant, (ii) quality-adjusted life-years (QALYs), and (iii) total costs.

RESULTS

We found no difference in survival advantage of PTA compared with standard of care (hazard ratio, 1.09; 95% confidence interval, 0.56-2.14). However, pancreas transplant ($172,823, 6.87 QALY) was cost-saving compared with standard of care ($232,897, 6.04 QALY) for type 1 diabetes. Pancreas transplantation was cost-effective in 95% of 10,000 simulations in probabilistic sensitivity analysis, using a $100,000/QALY willingness-to-pay threshold.

CONCLUSIONS

Although there is no difference in survival for PTA compared with standard of care, PTA is a cost-saving therapy for type 1 diabetes.

Automated Insulin Delivery in Real Life (AID-IRL): Real-World User Perspectives on Commercial AID

Commercial automated insulin delivery (AID) systems are usually assessed based on clinical outcomes, ignoring uptake. A qualitative study evaluated user experiences when switching to currently available commercial AID. Interview feedback was coded on key themes including the adoption experience with regards to quality of life, clinical outcomes, and users' expectations. Most felt their learning curve was easy. Most saw reduced hypoglycemia and increased time in range, although there were outliers. Many mentioned post-meal hyperglycemia as an improvement area for commercial AID. Users with one particular continuous glucose monitor (CGM) type reported sleep disruption. Companies should consider real-world user feedback with regards to improving training materials for new users with less CGM experience and by improving target flexibility and postprandial algorithm performance, plus reducing manual interventions required by users.