Personalised Insulin Adjustment with Reinforcement Learning: An In-Silico Validation for People with Diabetes on Intensive Insulin Treatment

2025-05-20

AI Summary

To address the challenge of insulin dose titration in patients with type 1 and long-standing type 2 diabetes, this paper proposes ABBA, a personalized basal-bolus insulin recommendation system based on reinforcement learning. ABBA introduces the first adaptive RL framework specifically designed for insulin therapy, integrating the UVA/Padova glucose kinetics model with the FDA-qualified virtual patient simulation platform and employing the Proximal Policy Optimization (PPO) algorithm to enable dynamic, continuous, and individualized dose optimization. Evaluated on 202 virtual patients, ABBA significantly improved time-in-range (TIR) while reducing time spent in hypoglycemia and hyperglycemia; performance consistently strengthened over two months and surpassed conventional titration protocols. The core contribution lies in the deep coupling of deep reinforcement learning with clinical insulin treatment pathways, yielding an interpretable, deployable, closed-loop dose optimization paradigm.

Technology Category

• Multiagent Systems: Adversarial Agents
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Despite recent advances in insulin preparations and technology, adjusting insulin remains an ongoing challenge for the majority of people with type 1 diabetes (T1D) and longstanding type 2 diabetes (T2D). In this study, we propose the Adaptive Basal-Bolus Advisor (ABBA), a personalised insulin treatment recommendation approach based on reinforcement learning for individuals with T1D and T2D, performing self-monitoring blood glucose measurements and multiple daily insulin injection therapy. We developed and evaluated the ability of ABBA to achieve better time-in-range (TIR) for individuals with T1D and T2D, compared to a standard basal-bolus advisor (BBA). The in-silico test was performed using an FDA-accepted population, including 101 simulated adults with T1D and 101 with T2D. An in-silico evaluation shows that ABBA significantly improved TIR and significantly reduced both times below- and above-range, compared to BBA. ABBA's performance continued to improve over two months, whereas BBA exhibited only modest changes. This personalised method for adjusting insulin has the potential to further optimise glycaemic control and support people with T1D and T2D in their daily self-management. Our results warrant ABBA to be trialed for the first time in humans.

Problem

• Personalized insulin adjustment for T1D and T2D patients
• Improving time-in-range (TIR) with reinforcement learning
• Comparing ABBA efficacy against standard basal-bolus advisor

Innovation

• Reinforcement learning for insulin adjustment
• Personalised Adaptive Basal-Bolus Advisor (ABBA)
• In-silico validation with FDA-accepted population