Artificial intelligence–based insulin delivery and decision-support systems produced glycemic outcomes comparable or superior to comparator approaches while reducing or not increasing hypoglycemia across five clinical trials, according to a systematic review presented at ENDO 2026, the Endocrine Society annual meeting, in Chicago.  

Although artificial intelligence (AI)-driven automated insulin delivery (AID) systems are increasingly used in diabetes care, quantitative synthesis of their efficacy and safety across clinical settings remains limited. To address this gap, researchers assessed the efficacy and safety of AI-based insulin delivery and decision-support systems in type 1 and type 2 diabetes. 

Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines, the investigators searched PubMed, Embase, and ClinicalTrials.gov and identified 96 records.  After screening and full-text review, five clinical trials evaluating AI-based insulin delivery or decision-support systems in patients with diabetes met the inclusion criteria. The outcomes extracted were time in range (TIR) between 70 and 180 mg/dL and hypoglycemia below 70 mg/dL. 

In one trial, among patients with type 1 diabetes during exercise, a hybrid machine learning model predictive control system increased TIR from 70% to 75% and reduced exercise-associated hypoglycemia. 

In another involving adults with type 1 diabetes, an artificial neural network–based controller achieved a TIR of 86% compared with 87% for a traditional model predictive control algorithm. Although overall glycemic control was similar, the neural network significantly reduced time below range while requiring sixfold less computation. 

A separate study evaluating a fuzzy logic–based overnight insulin control system reported improvement in TIR from 55% in the control condition to 76% with the AI system, with no severe hypoglycemic events reported. 

One study involved hospitalized patients with type 2 diabetes. In that setting, AI-guided insulin titration achieved a TIR of 76% compared with 74% under physician-led insulin management and produced fewer hypoglycemic episodes. 

Finally, among youth with type 1 diabetes receiving real-world care, an AI decision-support system produced a TIR of 50% compared with 52% under specialist management with equivalent safety and no increase in hypoglycemia. 

"Quantitative evidence from clinical trials indicates that AI-based AID and decision-support systems achieve TIRs of 50.2% to 86%, are non-inferior or superior to standard care, and consistently reduce or do not increase hypoglycemia," the investigators wrote. The findings "underscore the need for larger, longer-term trials to confirm durability and real-world impact as current practice integrates hybrid AI closed-loop systems and AI-guided insulin management with clinician and nursing oversight." 

The abstract did not report funding sources or conflicts of interest.