Effects of an artificial pancreas on postoperative inflammation in patients with esophageal cancer.

PURPOSES: Subtotal esophagectomy for esophageal cancer (EC) is associated with high morbidity rates. Tight glycemic control using an artificial pancreas (AP) is one of the promising strategies to reduce postoperative inflammation and morbidities. However, the effects of tight glycemic control using AP in patients with EC are yet to be fully elucidated. METHOD: This study reviewed 96 patients with EC who underwent subtotal esophagectomy. The postoperative inflammation parameters and morbidity rates were compared between patients who used the AP (n = 27) or not (control group, n = 69). AP is a closed-loop system that comprises a continuous glucose monitor and an insulin pump. RESULTS: The numbers of white blood cells (WBC) and Neutrophils (Neut) were noted to be lower in the AP group than in the control group, but with no significant difference. The ratio in which the number of WBC, Neut, and CRP on each postoperative day (POD) was divided by those tested preoperatively was used to standardize the results. The ratio of WBC and Neut on 1POD was significantly lower in the AP group than in the control group. The rate of surgical site infection was lower in the AP group than in the control group. CONCLUSION: AP significantly decreased WBC and Neut on 1POD; this suggests the beneficial effects of AP in alleviating postoperative inflammation.

An intelligent wearable artificial pancreas patch based on a microtube glucose sensor and an ultrasonic insulin pump.

In order to improve the living standards of diabetes patients and reduce the negative health effects of this disease, the medical community has been actively searching for more effective treatments. In recent years, an artificial pancreas has emerged as an important approach to managing diabetes. Despite these recent advances, meeting the requirements for miniaturized size, accurate sensing and large-volume pumping capability remains a great challenge. Here, we present a novel miniaturized artificial pancreas based on a long microtube sensor integrated with an ultrasonic pump. Our device meets the requirements of achieving both accurate sensing and high pumping capacity. The artificial pancreas is constructed based on a long microtube that is low cost, painless and simple to operate, where the exterior of the microtube is fabricated as a glucose sensor for detecting diabetes and the interior of the microtube is used as a channel for delivering insulin through an ultrasonic pump. This work successfully achieved closed-loop control of blood glucose and treatment of diabetes in rats. It is expected that this work can open up new methodologies for the development of microsystems, and advance the management approach for diabetes patients.

Study protocol for a randomised open-label clinical trial examining the safety and efficacy of the Android Artificial Pancreas System (AAPS) with advanced bolus-free features in adults with type 1 diabetes: the 'CLOSE IT' (Closed Loop Open SourcE In Type 1 diabetes) trial.

INTRODUCTION: Multiple automated insulin delivery (AID) systems have become commercially available following randomised controlled trials demonstrating benefits in people with type 1 diabetes (T1D). However, their real-world utility may be undermined by user-associated burdens, including the need to carbohydrate count and deliver manual insulin boluses. There is an important need for a 'fully automated closed loop' (FCL) AID system, without manual mealtime boluses. The (Closed Loop Open SourcE In Type 1 diabetes) trial is a randomised trial comparing an FCL AID system to the same system used as a hybrid closed loop (HCL) in people with T1D, in an outpatient setting over an extended time frame. METHODS AND ANALYSIS: Randomised, open-label, parallel, non-inferiority trial comparing the Android Artificial Pancreas System (AAPS) AID algorithm used as FCL to the same algorithm used as HCL. Seventy-five participants aged 18-70 will be randomised (1:1) to one of two treatment arms for 12 weeks: (a) FCL-participants will be advised not to bolus for meals and (b) HCL-participants will use the AAPS AID algorithm as HCL with announced meals. The primary outcome is the percentage of time in target sensor glucose range (3.9-10.0 mmol/L). Secondary outcomes include other glycaemic metrics, safety, psychosocial factors, platform performance and user dietary factors. Twenty FCL arm participants will participate in a 4-week extension phase comparing glycaemic and dietary outcomes using NovoRapid (insulin aspart) to Fiasp (insulin aspart and niacinamide). ETHICS AND DISSEMINATION: Approvals are by the Alfred Health Ethics Committee (615/22) (Australia) and Health and Disability Ethics Committees (2022 FULL 13832) (New Zealand). Each participant will provide written informed consent. Data protection and confidentiality will be ensured. Study results will be disseminated by publications, conferences and patient advocacy groups. TRIAL REGISTRATION NUMBERS: ACTRN12622001400752 and ACTRN12622001401741.

12-Month Efficacy of Advanced Hybrid Closed-Loop System in Adult Type 1 Diabetes Patients.

Automated insulin delivery (AID) systems have improved glycemic control in individuals with type 1 diabetes (T1D). The "advanced hybrid closed loop" (AHCL) stands out as the most recent development in AID systems for T1D management. In a real-world clinical environment, we retrospectively evaluated the AHCL MiniMed™ 780G system's effectiveness to achieve and sustain glycemic control over a 12-month period in 22 adult T1D subjects. Within just 14 days of activating the automatic mode, the AHCL MiniMed 780G system showed rapid improvement in glycemic control, which persisted for 12 months. These findings underscore the effectiveness of AHCL systems in achieving and preserving optimal glycemic control in adults with T1D over a very long follow-up.

Integration of a Safety Module to Prevent Rebound Hypoglycemia in Closed-Loop Artificial Pancreas Systems.

BACKGROUND: With automated insulin delivery (AID) systems becoming widely adopted in the management of type 1 diabetes, we have seen an increase in occurrences of rebound hypoglycemia or generated hypoglycemia induced by the controller's response to rapid glucose rises following rescue carbohydrates. Furthermore, as AID systems aim to enable complete automation of prandial control, algorithms are designed to react even more strongly to glycemic rises. This work introduces a rebound hypoglycemia prevention layer (HypoSafe) that can be easily integrated into any AID system. METHODS: HypoSafe constrains the maximum permissible insulin delivery dose based on the minimum glucose reading from the previous hour and the current glucose level. To demonstrate its efficacy, we integrated HypoSafe into the latest University of Virginia (UVA) AID system and simulated two scenarios using the 100-adult cohort of the UVA/Padova T1D simulator: a nominal case including three unannounced meals, and another case where hypoglycemia was purposely induced by an overestimated manual bolus. RESULTS: In both simulation scenarios, rebound hypoglycemia events were reduced with HypoSafe (nominal: from 39 to 0, hypo-induced: from 55 to 7) by attenuating the commanded basal (nominal: 0.27U vs. 0.04U, hypo-induced: 0.27U vs. 0.03U) and bolus (nominal: 1.02U vs. 0.05U, hypo-induced: 0.43U vs. 0.02U) within the 30-minute interval after treating a hypoglycemia event. No clinically significant changes resulted for time in the range of 70 to 180 mg/dL or above 180 mg/dL. CONCLUSION: HypoSafe was shown to be effective in reducing rebound hypoglycemia events without affecting achieved time in range when combined with an advanced AID system.

Effect of artificial pancreas system use on glycaemic control among pregnant women with type 1 diabetes mellitus: A meta-analysis of randomized controlled trials.

AIM: To assess the efficacy of artificial pancreas systems (APS) use among pregnant women with type 1 diabetes mellitus (T1DM) by conducting a meta-analysis. METHODS: We searched five databases, including EMBASE, Web of Science, PubMed, Cochrane Library and SCOPUS, for literature on APS use among pregnant women with T1DM before October 9, 2023. The primary endpoint was 24-hour time in range (TIR; 3.5-7.8 mmol/L). Secondary endpoints included glycaemic metrics for 24-hour (mean blood glucose [MBG], time above range [TAR], time below range [TBR]), and overnight TIR and TBR. RESULTS: We identified four randomized controlled trials involving 164 participants; one study with 16 participants focused on overnight APS use, and the other three focused on 24-hour APS use. Compared with standard care, APS exhibited a favourable effect on 24-hour TIR (standard mean difference [SMD] = 0.53, 95% confidence interval [CI] 0.25, 0.80, P < 0.001), overnight TIR (SMD = 0.67, 95% CI 0.39, 0.95, P < 0.001), and overnight TBR (<3.5 mmol/L; SMD = -0.49, 95% CI -0.77, -0.21 P < 0.001), while there was no significant difference in 24-hour TAR, 24-hour TBR, or MBG between the two groups. We further conducted subgroup analyses after removing the trial focused on overnight APS use and showed that 24-hour APS use reduced not only the 24-hour TIR (SMD = 0.41, 95% CI 0.12, 0.71; P = 0.007) but also the 24-hour TBR (<2.8 mmol/L; SMD = -0.77, 95% CI -1.32, -0.23, P = 0.006). CONCLUSION: Our findings suggest that APS might improve 24-hour TIR and overnight glycaemic control, and 24-hour APS use also significantly reduced 24-hour TBR (2.8 mmol/L) among pregnant women with T1DM.

Estimation and Prediction of Glucose Appearance Rate for Use in a Fully Closed-Loop Dual-Hormone Intraperitoneal Artificial Pancreas.

OBJECTIVE: A fully automated artificial pancreas requires a meal estimator and predictions of blood glucose levels (BGL) to handle disturbances during meal times, all without relying on manual meal announcements and user interventions. This study introduces a technique for estimating the glucose appearance rate (GAR) and predicting BGL in people with type 1 diabetes and insulin and glucagon administration. It is demonstrated for intraperitoneal insulin and glucagon delivery but may be adapted to other delivery sites. METHOD: The estimator is designed based on the moving horizon estimation (MHE) approach, where the underlying cost function incorporates prior statistical information on the GAR in subjects over the course of a day. The proposed prediction scheme is developed to predict GAR using estimated states and an intestinal model, which is then used to predict BGL with the help of an animal glucose metabolic model. RESULTS: The intraperitoneal dual-hormone estimator was evaluated on three anesthetized animals, achieving a 21.8% mean absolute percentage error (MAPE) for GAR estimation and a 10.0% MAPE for BGL prediction when the future GAR is known. For a 120-minute prediction horizon, the proposed predictor achieved an 18.0% MAPE for GAR and a 28.4% MAPE for BGL. CONCLUSION: The findings demonstrate the effectiveness and reliability of the proposed estimator and its potential for use in a fully automated artificial pancreas and reducing user interventions. SIGNIFICANCE: This study represents advancements toward the development of a fully automated artificial pancreas, ultimately enhancing the quality of life for people with type 1 diabetes.

The artificial pancreas: two alternative approaches to achieve a fully closed-loop system with optimal glucose control.

INTRODUCTION: Diabetes mellitus type 1 is a chronic disease that implies mandatory external insulin delivery. The patients must monitor their blood glucose levels and administer appropriate insulin boluses to keep their blood glucose within the desired range. It requires a lot of time and endeavour, and many patients struggle with suboptimal glucose control despite all their efforts. MATERIALS AND METHODS: This narrative review combines existing knowledge with new discoveries from animal experiments. DISCUSSION: In the last decade, artificial pancreas (AP) devices have been developed to improve glucose control and relieve patients of the constant burden of managing their disease. However, a feasible and fully automated AP is yet to be developed. The main challenges preventing the development of a true, subcutaneous (SC) AP system are the slow dynamics of SC glucose sensing and particularly the delay in effect on glucose levels after SC insulin infusions. We have previously published studies on using the intraperitoneal space for an AP; however, we further propose a novel and potentially disruptive way to utilize the vasodilative properties of glucagon in SC AP systems. CONCLUSION: This narrative review presents two lesser-explored viable solutions for AP systems and discusses the potential for improvement toward a fully automated system: A) using the intraperitoneal approach for more rapid insulin absorption, and B) besides using glucagon to treat and prevent hypoglycemia, also administering micro-boluses of glucagon to increase the local SC blood flow, thereby accelerating SC insulin absorption and SC glucose sensor site dynamics.

Open-source Artificial Pancreas Systems Are Safe and Effective When Supported In-clinic: Outcomes in 248 Consecutive Type 1 Diabetes Clients.

OBJECTIVE: Our aim in this study was to determine the safety, glycemia, and quality of life (QoL) associated with in-clinic installation and management of supported open-source artificial pancreas systems (SOSAPS) in type 1 diabetes (T1D). METHODS: This investigation is a retrospective cohort study of consecutive SOSAPS users at a Canadian diabetes centre. SOSAPS were offered to all moderately tech-savvy T1D clients on sensor-augmented multiple daily injection or pump, able to pay for hardware, and willing to sign a consent and waiver document. SOSAPS were installed and maintained by clinic staff at no cost to clients. iPhone users were assigned to either Loop (n=108) or iPhone artificial pancreas systems (iAPS; n=114) and Android users to Android-type APS (n=24). Outcomes included severe hypoglycemia and diabetic ketoacidosis (DKA), time in range (TIR) 4.0 to 10.0 mmol/L, time below range (TBR) <4 mmol/L, glucose management indicator (GMI), mean sensor glucose (MSG), change in glycated hemoglobin (A1C), and QoL. RESULTS: Two hundred forty-eight subjects (131 males, 117 females), with a mean age of 36 years and diabetes duration of 21 years, experienced 3 episodes of severe hypoglycemia and no DKA over a follow-up of 17 months. TIR rose by 16%, from 64% to 80% (p<0.0001); TBR fell by 1.0%, from 3.5% to 2.5% (p=0.001); MSG fell from 9.0 to 8.1 mmol/L (p<0.001); GMI fell from 7.3% to 6.7% (p<0.001); and A1C fell from 7.2% to 6.7% (p<0.0001). QoL scores were healthy before and improved after SOSAPS. CONCLUSIONS: Clients with T1D using SOSAPS and supported with no-cost care to the client (software, technology, and physician/physician assistant) safely achieved improved TIR, GMI, A1C, and QoL.

NIH Fifth Artificial Pancreas Workshop 2023: Meeting Report: The Fifth Artificial Pancreas Workshop: Enabling Fully Automation, Access, and Adoption.

The Fifth Artificial Pancreas Workshop: Enabling Fully Automation, Access, and Adoption was held at the National Institutes of Health (NIH) Campus in Bethesda, Maryland on May 1 to 2, 2023. The organizing Committee included representatives of NIH, the US Food and Drug Administration (FDA), Diabetes Technology Society, Juvenile Diabetes Research Foundation (JDRF), and the Leona M. and Harry B. Helmsley Charitable Trust. In previous years, the NIH Division of Diabetes, Endocrinology, and Metabolic Diseases along with other diabetes organizations had organized periodic workshops, and it had been seven years since the NIH hosted the Fourth Artificial Pancreas in July 2016. Since then, significant improvements in insulin delivery have occurred. Several automated insulin delivery (AID) systems are now commercially available. The workshop featured sessions on: (1) Lessons Learned from Recent Advanced Clinical Trials and Real-World Data Analysis, (2) Interoperability, Data Management, Integration of Systems, and Cybersecurity, Challenges and Regulatory Considerations, (3) Adaptation of Systems Through the Lifespan and Special Populations: Are Specific Algorithms Needed, (4) Development of Adaptive Algorithms for Insulin Only and for Multihormonal Systems or Combination with Adjuvant Therapies and Drugs: Clinical Expected Outcomes and Public Health Impact, (5) Novel Artificial Intelligence Strategies to Develop Smarter, More Automated, Personalized Diabetes Management Systems, (6) Novel Sensing Strategies, Hormone Formulations and Delivery to Optimize Close-loop Systems, (7) Special Topic: Clinical and Real-world Viability of IP-IP Systems. "Fully automated closed-loop insulin delivery using the IP route," (8) Round-table Panel: Closed-loop performance: What to Expect and What are the Best Metrics to Assess it, and (9) Round-table Discussion: What is Needed for More Adaptable, Accessible, and Usable Future Generation of Systems? How to Promote Equitable Innovation? This article summarizes the discussions of the Workshop.

InsNET: Accurate Basal and Bolus Insulin Dose Prediction for Closed Loop Diabetes Management.

It has been demonstrated that closed-loop diabetes management results in better glycemic control and greater compliance than open-loop diabetes management. Deep learning models have been used to implement different components of artifical pancreas. In this work, a novel deep learning model InsNET has been proposed to estimate the basal and bolus insulin level and insulin bolus in patients with type I diabetes utilizing subcutaneous insulin infusion pumps for closed loop diabetes management system. The proposed InsNET is formed with a Wide-Deep combination of LSTM and GRU layers. Additionally, physical activity level has been included as an input in comparison to previous models where only past glucose levels (CGM), meal intake (CHO) and past insulin dosage were used as inputs. The proposed model was tested on In-silico data, and it achieved a Mean Absolute Error (MAE) of 0.002 and Root Mean Squared Error (RMSE) of 0.007 for UVA/Padova Dataset and MAE of 0.001 and RMSE OF 0.003 for mGIPsim Dataset.Clinical relevance- Insulin dose determination is an important as aspect of artificial pancreas. This work describes a deep learning model to determine accurate basal and bolus insulin dosage.

Controlling and powering a fully implantable artificial pancreas refillable by ingestible pills.

Over the past decade, there has been a growing interest in the development of an artificial pancreas for intraperitoneal insulin delivery. Intraperitoneal implantable pumps guarantee more physiological glycemic control than subcutaneous wearable ones, for the treatment of type 1 diabetes. In this work, a fully implantable artificial pancreas refillable by ingestible pills is presented. In particular, solutions enabling the communication between the implanted pump and external user interfaces and novel control algorithms to intraperitoneally release an adequate amount of insulin based on glycemic data are shown. In addition, the powering and the wireless battery recharging are addressed. Specifically, the design and optimization of a customized transcutaneous energy transfer with two independent wireless channels are presented. The system was tested in terms of recharging efficacy, possible temperature rise within the body, during the recharging process and reliability of the wireless connection in the air and in the presence of ex vivo tissues.Clinical Relevance- This work aims to improve the control, battery recharging, and wireless communication of a fully implantable artificial pancreas for type 1 diabetes treatment.

Multivariable Automated Insulin Delivery System for Handling Planned and Spontaneous Physical Activities.

BACKGROUND: Hybrid closed-loop control of glucose levels in people with type 1 diabetes mellitus (T1D) is limited by the requirements on users to manually announce physical activity (PA) and meals to the artificial pancreas system. Multivariable automated insulin delivery (mvAID) systems that can handle unannounced PAs and meals without any manual announcements by the user can improve glycemic control by modulating insulin dosing in response to the occurrence and intensity of spontaneous physical activities. METHODS: An mvAID system is developed to supplement the glucose measurements with additional physiological signals from a wristband device, with the signals analyzed using artificial intelligence algorithms to automatically detect the occurrence of PA and estimate its intensity. This additional information gained from the physiological signals enables more proactive insulin dosing adjustments in response to both planned exercise and spontaneous unanticipated physical activities. RESULTS: In silico studies of the mvAID illustrate the safety and efficacy of the system. The mvAID is translated to pilot clinical studies to assess its performance, and the clinical experiments demonstrate an increased time in range and reduced risk of hypoglycemia following unannounced PA and meals. CONCLUSIONS: The mvAID systems can increase the safety and efficacy of insulin delivery in the presence of unannounced physical activities and meals, leading to improved lives and less burden on people with T1D.

Clinical Translation of Bio-Artificial Pancreas Therapies: Ethical, Legal and Psychosocial Interdisciplinary Considerations and Key Recommendations.

The field of regenerative medicine offers potential therapies for Type 1 Diabetes, whereby metabolically active cellular components are combined with synthetic medical devices. These therapies are sometimes referred to as "bioartificial pancreases." For these emerging and rapidly developing therapies to be clinically translated to patients, researchers must overcome not just scientific hurdles, but also navigate complex legal, ethical and psychosocial issues. In this article, we first provide an introductory overview of the key legal, ethical and psychosocial considerations identified in the existing literature and identify areas where research is currently lacking. We then highlight two principal areas of concern in which these discrete disciplines significantly overlap: 1) individual autonomy and 2) access and equality. Using the example of beta-cell provenance, we demonstrate how, by harnessing an interdisciplinary approach we can address these key areas of concern. Moreover, we provide practical recommendations to researchers, clinicians, and policymakers which will help to facilitate the clinical translation of this cutting-edge technology for Type 1 Diabetes patients. Finally, we emphasize the importance of exploring patient perspectives to ensure their responsible and acceptable translation from bench to body.

Efficacy of automated insulin delivery (AID) systems in type 1 diabetes: protocol of a systematic review and network meta-analysis of outpatient randomised controlled trials.

INTRODUCTION: Automated insulin delivery (AID), also known as artificial pancreas system or 'closed-loop system', represents a novel option for current treatments for type 1 diabetes (T1D). The objective of this systematic review and meta-analysis is to assess the efficacy of AID systems in comparison with current intensified insulin therapy for glycaemic control and patient-reported outcomes in individuals with T1D. METHODS AND ANALYSIS: Studies will be eligible if they are randomised controlled trials (RCTs) in people with T1D of all ages, and if they compare an AID system for self-administration during the day and night period with any other type of insulin therapy for at least 3 weeks. The primary outcome will be time in the glucose target range of 70-180 mg/dL. A systematic review will be conducted in the MEDLINE, Embase, Cochrane Central Register of Controlled Trials and ClinicalTrials.gov registries from their inception dates. Two authors will independently screen all references based on titles and abstracts against the eligibility criteria. For data extraction, standard forms will be developed and tested before extraction. All information will be assessed independently by at least two reviewers. The risk of bias of the included studies will be assessed using the Cochrane Risk of Bias 2 tool. The data synthesis will include a random-effects pairwise and network meta-analysis (NMA) in a frequentist framework. Where applicable and if sufficient RCTs are available, sensitivity analyses will be performed, and heterogeneity and publication bias will be assessed. The certainty of evidence from the NMA will be evaluated following the Grading of Recommendations Assessment, Development, and Evaluation working group guidance. ETHICS AND DISSEMINATION: No ethical approval is needed. The results will be reported to the funder, presented in a peer-reviewed scientific journal and at conferences, and disseminated via press release, social media and public events. PROSPERO REGISTRATION NUMBER: CRD42023395492.

Developing the UVA/Padova Type 1 Diabetes Simulator: Modeling, Validation, Refinements, and Utility.

Arguably, diabetes mellitus is one of the best quantified human conditions. In the past 50 years, the metabolic monitoring technologies progressed from occasional assessment of average glycemia via HbA1c, through episodic blood glucose readings, to continuous glucose monitoring (CGM) producing data points every few minutes. The high-temporal resolution of CGM data enabled increasingly intensive treatments, from decision support assisting insulin injection or oral medication, to automated closed-loop control, known as the "artificial pancreas." Throughout this progress, mathematical models and computer simulation of the human metabolic system became indispensable for the technological progress of diabetes treatment, enabling every step, from assessment of insulin sensitivity via the now classic Minimal Model of Glucose Kinetics, to in silico trials replacing animal experiments, to automated insulin delivery algorithms. In this review, we follow these developments, beginning with the Minimal Model, which evolved through the years to become large and comprehensive and trigger a paradigm change in the design of diabetes optimization strategies: in 2007, we introduced a sophisticated model of glucose-insulin dynamics and a computer simulator equipped with a "population" of N = 300 in silico "subjects" with type 1 diabetes. In January 2008, in an unprecedented decision, the Food and Drug Administration (FDA) accepted this simulator as a substitute to animal trials for the pre-clinical testing of insulin treatment strategies. This opened the field for rapid and cost-effective development and pre-clinical testing of new treatment approaches, which continues today. Meanwhile, animal experiments for the purpose of designing new insulin treatment algorithms have been abandoned.

Efficacy and safety of Android artificial pancreas system use at home among adults with type 1 diabetes mellitus in China: protocol of a 26-week, free-living, randomised, open-label, two-arm, two-phase, crossover trial.

INTRODUCTION: Do-it-yourself artificial pancreas system (DIY APS) is built using commercially available insulin pump, continuous glucose monitoring (CGM) and an open-source algorithm. Compared with commercial products, DIY systems are affordable, allow personalised settings and provide updated algorithms, making them a more promising therapy for most patients with type 1 diabetes mellitus (T1DM). Many small and self-reported observational studies have found that their real-world use was associated with potential metabolic and psychological benefits. However, rigorous-designed studies are urgently needed to confirm its efficacy and safety. METHODS AND ANALYSIS: In this 26-week randomised, open-label, two-arm, two-phase, crossover trial, participants aged 18-75 years, with T1DM and glycated haemoglobin (HbA1c) 7-11%, will use AndroidAPS during one 12-week period and sensor-augmented pump during another 12-week period. This study will recruit at least 24 randomised participants. AndroidAPS consists of three components: (1) real-time CGM; (2) insulin pump; (3) AndroidAPS algorithm implemented in Android smartphone. The primary endpoint is time in range (3.9-10.0 mmol/L) derived from CGM. The main secondary endpoints include percentage of sensor glucose values below, within and above target range; mean sensor glucose value; measures of glycaemic variability and centralised HbA1c. Safety endpoints mainly include the frequency of hypoglycaemia events, diabetic ketoacidosis and other serious adverse events. ETHICS AND DISSEMINATION: This study has been approved by the Ethics Committee of the Third Affiliated Hospital of Sun Yat-sen University. There will be verbal and written information regarding the trial given to each participant. The study will be disseminated through peer-reviewed publications and conference presentations. OVERALL STATUS: Recruiting. STUDY START: 11 February 2023. PRIMARY COMPLETION: 31 July 2024. TRIAL REGISTRATION NUMBER: ClinicalTrials.gov Registry (NCT05726461).

Offline Deep Reinforcement Learning and Off-Policy Evaluation for Personalized Basal Insulin Control in Type 1 Diabetes.

Recent advancements in hybrid closed-loop systems, also known as the artificial pancreas (AP), have been shown to optimize glucose control and reduce the self-management burdens for people living with type 1 diabetes (T1D). AP systems can adjust the basal infusion rates of insulin pumps, facilitated by real-time communication with continuous glucose monitoring. Deep reinforcement learning (DRL) has introduced new paradigms of basal insulin control algorithms. However, all the existing DRL-based AP controllers require extensive random online interactions between the agent and environment. While this can be validated in T1D simulators, it becomes impractical in real-world clinical settings. To this end, we propose an offline DRL framework that can develop and validate models for basal insulin control entirely offline. It comprises a DRL model based on the twin delayed deep deterministic policy gradient and behavior cloning, as well as off-policy evaluation (OPE) using fitted Q evaluation. We evaluated the proposed framework on an in silico dataset generated by the UVA/Padova T1D simulator, and the OhioT1DM dataset, a real clinical dataset. The performance on the in silico dataset shows that the offline DRL algorithm significantly increased time in range while reducing time below range and time above range for both adult and adolescent groups. Then, we used the OPE to estimate model performance on the clinical dataset, where a notable increase in policy values was observed for each subject. The results demonstrate that the proposed framework is a viable and safe method for improving personalized basal insulin control in T1D.