Artificial intelligence based clinical decision systems are safe and effective for diabetes management

1. In this randomized clinical trial of adults with type 2 diabetes (T2D), the artificial intelligence-based insulin clinical decision support system (iNCDSS) was noninferior to standard physician-administered insulin therapy.

2. Compared to the physician group, there were no significant differences in adverse events in the iNCDSS group. 

Evidence Rating Level: 1 (Excellent) 

Study Rundown: Type 2 diabetes (T2D) is one of the most commonly encountered diseases globally, with considerable morbidity and mortality rates. Many adults with T2D are treated with insulin, however, less than one-third reach their target sugar goals. Some barriers include a lack of access to endocrinologists, fear of hypoglycemia, and prescription complexity. As a result, different digital strategies have emerged as possible treatment options for glucose control. Artificial intelligence has been increasingly used in healthcare. This study sought to assess the efficacy and safety of the artificial intelligence-based insulin clinical decision support system (iNCDSS) in patients with T2D receiving insulin therapy. To achieve this, a randomized clinical trial was conducted comparing insulin titration guided by iNCDSS with standard care provided by senior endocrinologists over a 5-day inpatient period. The main outcomes of interest included the proportion of time in the target glucose range, the proportion of time when glucose was above or below the target range, along with others. On average, individuals in the iNCDSS group stayed within the target glucose range for slightly longer than individuals in the physician group, proving the effectiveness of the AI-assisted dosing. The safety of iNCDSS was also supported as there were low rates of hypoglycemia and hyperglycemia, and no significant differences in adverse events between the groups. Since the study was only conducted in Chinese hospitals, the generalizability of the results may be limited. Despite this, the randomized clinical trial provides evidence supporting that iNCDSS was a safe and effective way to monitor insulin dosage in individuals with type 2 diabetes. Optimal insulin therapy 

Click here to read the study in JAMA Network Open

Relevant reading: Integrating AI in Diabetes Treatment. 

In-depth [randomized controlled trial]: A multicenter randomized clinical trial was conducted to best assess the efficacy and safety of AI-assisted insulin therapy. Participants were eligible if they were aged 18 years or older, diagnosed with T2D with HbA1c levels between 7.0% and 11.0%, and were treated with diet, oral antidiabetic agents and/or insulin in the previous 3 months. After meeting eligibility, the participants were randomized 1:1 to receive insulin by iNCDSS or endocrinology physicians. Flash glucose monitoring (Freestyle Libre) was used to conduct continuous glucose monitoring (CGM) in both groups. A total of 149 individuals participated in the study (mean [SD] age 64.2 [12.0] years) and were randomized, with 75 in the iNCDSS group and 74 in the physician group. The mean (SD) target glucose range (TIR) was 76.4% (16.4%) in the intervention group and 73.6% (16.8%) in the control group. During the continuous glucose monitoring, there were no significant differences between the groups for glucose levels. The iNCDSS group had a significantly lower daily insulin dose (27.2 [20.0-35.5] units) compared to the physician group (30.2 [22.8-42.6] units; P=.01). However, baseline insulin dose was considered, there was no statistically significant difference, suggesting the starting insulin doses was the main cause of the previous difference. Between the groups, there were no significant differences in adverse events. Overall, in this randomized clinical trial, iNCDSS was not inferior to experienced endocrinologists in managing insulin dosing for T2D patients in the hospital. 

Image: PD

©2025 2 Minute Medicine, Inc. All rights reserved. No works may be reproduced without expressed written consent from 2 Minute Medicine, Inc. Inquire about licensing here. No article should be construed as medical advice and is not intended as such by the authors or by 2 Minute Medicine, Inc.