AI Tool Matches Senior Endocrinologists in Insulin Titration Accuracy: JAMA

The randomized clinical trial, published in JAMA Network Open, revealed that the recommendations of the AI tool were adopted in 99% of cases, and senior endocrinologists rated it with a high satisfaction score of 4.1 out of 5.0.

The trial, conducted across three medical centers between October 2021 and September 2022, involved 149 adult participants with T2D. All patients had HbA1c levels between 7.0% and 11.0% and had been on antidiabetic treatments in the preceding three months. Zhen Ying, Fudan University, Shanghai, China, and colleagues aimed to evaluate whether the AI-driven insulin clinical decision support system (iNCDSS) could achieve glycemic control outcomes comparable to those managed by seasoned physicians.

Participants were randomly assigned to receive insulin titration guidance either from the AI system or from senior endocrinologists over five consecutive days. The primary outcome measured was the percentage of time patients’ glucose levels remained within the target range (70–180 mg/dL) during the study period. A noninferiority margin of 6 percentage points was set as the benchmark.

The following were the key findings of the study:

• Patients in the AI-assisted group maintained blood glucose within the target range for an average of 76.4% of the time, compared to 73.6% in the physician-managed group.
• The observed difference of 2.7 percentage points fulfilled the trial’s predefined noninferiority margin.
• No significant increase in adverse events, such as hypoglycemia, was reported in the AI group, indicating that patient safety was not compromised.
• The findings highlight the potential of AI to provide real-time, personalized insulin management effectively.
• The AI system may help overcome current challenges in timely and individualized insulin dose adjustments in clinical practice.
• Participating physicians appreciated the tool for being time-saving and providing clear, efficient support in clinical decision-making.

"Given the growing global burden of type 2 diabetes and the challenges involved in precise insulin management, especially in inpatient care, these results significantly advance," the researchers note, adding that, "The AI tool could offer a scalable and efficient solution to enhance diabetes care, particularly in settings where access to endocrinologists is limited."

"Overall, the study demonstrates that AI-based clinical tools like the iNCDSS can provide a level of care that rivals that of experienced physicians. As digital health continues to evolve, such innovations are promising for chronic disease management by improving accessibility, efficiency, and consistency in care delivery," they concluded.

Reference:

Ying Z, Fan Y, Chen C, et al. Real-Time AI-Assisted Insulin Titration System for Glucose Control in Patients With Type 2 Diabetes: A Randomized Clinical Trial. JAMA Netw Open. 2025;8(5):e258910. doi:10.1001/jamanetworkopen.2025.8910