Objective: Using a multistep machine-learning procedure, add virtual continuous glucose monitoring (CGM) traces to the original sparse data of the landmark Diabetes Control and Complications Trial (DCCT). Assess the association of CGM metrics with the microvascular complications of type 1 diabetes observed during the DCCT and establish time-in-range (TIR) as a viable marker of glycemic control. Research Design and Methods: Utilizing the DCCT glycated hemoglobin data obtained every 1 or 3 months plus quarterly 7-point blood glucose (BG) profiles in a multistep procedure: (i) utilized archival BG traces to model interday BG variability and estimate glycated hemoglobin; (ii) trained across the DCCT BG profiles and associated each profile with an archival BG trace; and (iii) used previously identified CGM "motifs" to associate a CGM trace to a BG trace, for each DCCT participant. Results: TIR (70-180 mg/dL) computed from virtual CGM data over 14 days prior to each glycated hemoglobin measurement reproduced the observed glycemic control differences between the intensive and conventional DCCT groups, with TIR generally >60% and <40% in these groups, respectively. Similar to glycated hemoglobin, TIR was associated with the risk of development or progression of retinopathy, nephropathy, and neuropathy (all P-values <0.0001). Poisson regressions indicated that TIR predicted retinopathy and microalbuminuria similarly to the original glycated hemoglobin data. Conclusions: The landmark DCCT was revisited using contemporary data science methods, which allowed adding individual CGM traces to the original data. Fourteen-day CGM metrics predicted microvascular diabetes complications similarly to glycated hemoglobin. Clinical Trials Registration: Not a clinical trial.

The artificial pancreas is a closed-loop insulin delivery system that automatically regulates glucose levels in individuals with type 1 diabetes. In-silico testing using simulation environments accelerates the development of better artificial pancreas systems. Simulation environments need an accurate model that captures glucose dynamics during exercise to simulate real-life scenarios. We proposed six variations of the Bergman Minimal Model to capture the physiological effects of moderate exercise on glucose dynamics in individuals with type 1 diabetes. We estimated the parameters of each model with clinical data using a Bayesian approach and Markov chain Monte Carlo methods. The data consisted of measurements of plasma glucose, plasma insulin, and oxygen consumption collected from a study of 17 adults with type 1 diabetes undergoing aerobic exercise sessions. We compared the models based on the physiological plausibility of their parameters estimates and the deviance information criterion. The best model features (i) an increase in glucose effectiveness proportional to exercise intensity, and (ii) an increase in insulin action proportional to exercise intensity and duration. We validated the selected model by reproducing results from two previous clinical studies. The selected model accurately simulates the physiological effects of moderate exercise on glucose dynamics in individuals with type 1 diabetes. This work offers an important tool to develop strategies for exercise management with the artificial pancreas.

Objective: Multiple daily injections (MDI) therapy is the most common treatment for type 1 diabetes (T1D) including basal insulin doses to keep glucose levels constant during fasting conditions and bolus insulin doses with meals. Optimal insulin dosing is critical to achieving satisfactory glycemia but is challenging due to inter- and intra-individual variability. Here, we present a novel model-based iterative algorithm that optimizes insulin doses using previous-day glucose, insulin, and meal data. Methods: Our algorithm employs a maximum-a-posteriori method to estimate parameters of a model describing the effects of changes in basal-bolus insulin doses. Then, parameter estimates, their confidence intervals, and the goodness of fit, are combined to generate new recommendations. We assessed our algorithm in three ways. First, a clinical data set of 150 days (15 participants) was used to evaluate the proposed model and the estimation method. Second, a 60-day simulation was performed to demonstrate the efficacy of the algorithm. Third, a sample 6-day clinical experiment is presented and discussed. Results: The model fitted the clinical data well with a root-mean-square-error of 1.75 mmol/L. Simulation results showed an improvement in the time in target (3.9 10 mmol/L) from 64% to 77% and a decrease in the time in hypoglycemia (< 3.9 mmol/L) from 8.1% to 3.8%. The clinical experiment demonstrated the feasibility of the algorithm. Conclusion: Our algorithm has the potential to improve glycemic control in people with T1D using MDI. Significance: This work is a step forward towards a decision support system that improves their quality of life.

In this paper, we test and evaluate policy gradient reinforcement learning for automated blood glucose control in patients with Type 1 Diabetes Mellitus. Recent research has shown that reinforcement learning is a promising approach to accommodate the need for individualized blood glucose level control algorithms. The motivation for using policy gradient algorithms comes from the fact that adaptively administering insulin is an inherently continuous task. Policy gradient algorithms are known to be superior in continuous high-dimensional control tasks. Previously, most of the approaches for automated blood glucose control using reinforcement learning has used a finite set of actions. We use the Trust-Region Policy Optimization algorithm in this work. It represents the state of the art for deep policy gradient algorithms. The experiments are carried out in-silico using the Hovorka model, and stochastic behavior is modeled through simulated carbohydrate counting errors to illustrate the full potential of the framework. Furthermore, we use a model-free approach where no prior information about the patient is given to the algorithm. Our experiments show that the reinforcement learning agent is able to compete with and sometimes outperform state-of-the-art model predictive control in blood glucose regulation.