Time for Using Machine Learning for Dose Guidance in Titration of People With Type 2 Diabetes? A Systematic Review of Basal Insulin Dose Guidance

Abstract

Background: Real-world studies of people with type 2 diabetes (T2D) have shown insufficient dose adjustment during basal insulin titration in clinical practice leading to suboptimal treatment. Thus, 60% of people with T2D treated with insulin do not reach glycemic targets. This emphasizes a need for methods supporting efficient and individualized basal insulin titration of people with T2D. However, no systematic review of basal insulin dose guidance for people with T2D has been found. Objective: To provide an overview of basal insulin dose guidance methods that support titration of people with T2D and categorize these methods by characteristics, effect, and user experience. Methods: The review was conducted according to the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) guidelines. Studies about basal insulin dose guidance, including adults with T2D on basal insulin analogs published before September 7, 2022, were included. Joanna Briggs Institute critical appraisal checklists were applied to assess risk of bias. Results: In total, 35 studies were included, and three categories of dose guidance were identified: paper-based titration algorithms, telehealth solutions, and mathematical models. Heterogeneous reporting of glycemic outcomes challenged comparison of effect between the three categories. Few studies assessed user experience. Conclusions: Studies mainly used titration algorithms to titrate basal insulin as telehealth or in paper format, except for studies using mathematical models. A numerically larger proportion of participants seemed to reach target using telehealth solutions compared to paper-based titration algorithms. Exploring capabilities of machine learning may provide insights that could pioneer future research while focusing on holistic development.

Optimized glycemic control of type 2 diabetes with reinforcement learning: a proof-of-concept trial

“…Although some algorithms have been developed to assist physicians in insulin titration, only a few have been validated in clinical trials 31 , 32 . We conducted a proof-of-concept feasibility trial demonstrating the viability of the RL-DITR system in inpatients with T2D.…”

Intermittent Control for Safe Long-Acting Insulin Intensification for Type 2 Diabetes: In-Silico Experiments *

“…The authors noted that testing their method on a more detailed physiological model was warranted. Other proprietary automated devicesupported titration algorithms have been successfully evaluated in clinical trials [14]- [17], showing a trend in moving from paper-based SoC rules towards digital solutions [18]. Additionally, other strategies in people with T1D may be extended to T2D exist and are clinically validated [19], [20], or under ongoing clinical investigation [21], [22].…”