1,721,039 research outputs found
In silico testing of artificial pancreas and new type 1 diabetes treatments: model development and assessment
No full textIn healthy subjects, glucose regulation relies on a complex hormonal control system that maintains the blood glucose level in a safe range. Impairment of this regulatory system is the cause of several metabolic disorders, such as type 1 diabetes (T1DM), characterized by the absolute deficiency of insulin production, leading to a chronic hyperglycemia that, if not treated, can result in severe microvascular and macrovascular complications.
Currently, the best therapy for T1DM management makes use of a continuous subcutaneous insulin pump (CSII) coupled with a subcutaneous continuous glucose monitoring sensor (CGM), the so-called sensor-augmented-pump therapy (SAP). Nevertheless, to ease T1DM subject's life condition, in the last decade, the researchers have been focused on developing an automatic closed-loop control system for insulin infusion, the so-called Artificial Pancreas (AP), which aims to maintain the glucose level within the euglycemic range.
In this regards, simulation models allowed important steps forward in the AP research, enabling the possibility to perform several in silico tests, with relevant time- and cost- savings. In particular, in 2008 the US Food and Drug Administration accepted the T1DM simulator developed by Universities of Virginia (UVA) and Padova as a substitute for preclinical trials for certain insulin treatments, including closed-loop algorithms. This dramatically accelerated the process for the approval of human trials.
The UVA/Padova simulator (S2008) is based on a rather complex model of glucose dynamics that was identified on a data set of 204 healthy subjects for which not only plasma glucose and insulin measurements but also estimates of glucose fluxes were available. The simulator, equipped with 100 in silico adults, 100 adolescents, and 100 children, spanning the variability observed in the real type 1 diabetic population, has been updated in 2013 in order to better describe the distribution of glucose concentration observed in clinical trials (S2013). However, at the beginning of this project, the simulator validity was never been validated against clinical data. In addition, nowadays, the frontier of the AP research is the development of control algorithm effective for weekly or monthly use. However, the T1DM simulator was not fully adequate for the long-term testing, since its domain of validity was limited to a single-meal scenario.
The first aim of this research is thus to assess the simulator validity using data of the available clinical trials. The second objective is to extend the domain of validity of the simulator, making it suitable for simulating long-term clinical trials. Finally, a third scope is to illustrate the possible uses of the simulator, including setting up a paradigm for in silico trials for testing of new insulin treatments.
To achieve the first objective, a database of 24 T1DM subjects was first considered, who received dinner and breakfast in two occasions, for a total of 96 post-prandial glucose traces. Measured plasma glucose profiles were compared with those obtained with both S2008 and S2013, by replicating in 100 in silico adults the same experimental condition of the data (i.e. same meal amount and insulin delivery). The Continuous Glucose-Error Grid Analysis was used to assess the validity of the simulated traces, and the most common clinical outcome metrics, obtained in silico, were compared with the experimental ones. The results were satisfactory, proving that the virtual adults of the S2013 are representative of an age-matched T1DM population observed in a clinical trial.
Then, the T1DM model has been validated on 47 T1DM subjects who received dinner, breakfast and lunch, in three admissions, for a total of 23 hours per session. In particular, given the complexity of the model and the availability of glucose and insulin measurements only, a Bayesian approach has been adopted for model identification, considering, as \prior information, the parameter distribution included in the simulator for the generation of in silico subjects. Variability of model parameters describing glucose absorption and insulin sensitivity (SI, i.e. the ability of insulin to stimulate glucose disposal and suppress endogenous glucose production) was allowed, assuming that meal composition may be different at breakfast, lunch, and dinner (resulting in different absorption rate), and that SI may vary throughout the day. The model well described glucose traces and the posterior distribution of model parameters was similar to that included in the simulator; absorption parameters at breakfast were significantly different from those at lunch and dinner, reflecting more rapid dynamics of glucose absorption; on the other hand, insulin sensitivity varies in each individual but without a specific pattern. These results suggested the need of a time-varying simulator to better describe the glucose variability in the long-term.
In this regard, a model of intra-day variability of insulin sensitivity has been developed, by using data of a recent multiple tracer experiment performed in 20 T1DM subjects, which revealed the existence of diurnal patterns of SI, with SI lower at breakfast than lunch and dinner, on average. This difference was not statistically significant, both due to the small population size and the high inter-subject variability. In particular, seven SI daily patterns were identified, and their probabilities were estimated from the data. This information has been translated into the simulator by associating each in silico subject to one of the seven variability patterns, and modeling its SI with time-varying parameters. To test the goodness of the model, the same experimental protocol of the 20 T1DM subjects was replicated in silico: the comparison of simulated glucose against the data was satisfactory, showing that the simulated plasma glucose level was higher at breakfast than lunch and dinner, as the clinical data.
Finally, two case studies, illustrating the simulator employment, are presented. An AP adaptive control algorithm was tested first. The performance was evaluated in silico in a 1-month scenario: in particular, being able to provide a realistic time-varying behavior of the system, it was possible to prove evidence of the improved glucose control achievable with the adaptive control with respect to a non-adaptive AP algorithm. Then, the simulator was employed to evaluate the pharmacological effect of a novel inhaled insulin: in particular, the simulator have served to evaluate the post-prandial glucose in response to different insulin dosing regimens, thus allowing to determine, for each in silico subject, the best insulin pattern to optimally control post-prandial glucose.
In conclusion, in this work, the UVA/Padova T1DM simulator was assessed against clinical T1DM data; then, the T1DM model was identified on a 24-hour scenario, proving that a time-varying model was required to well describe the daily glucose variability; a model of intra-day variability of SI was thus developed and incorporated into the simulator. Finally, the use of the simulator was illustrated in two examples, i.e. the preclinical testing of an adaptive AP algorithm and the design of optimal dosing regimen of a novel inhaled insulin. In both cases, the simulator proved to be a useful tool for the in silico testing of T1DM treatments
In Silico Optimization of Long-Acting Insulin Injection Time in Subjects With Type 1 Diabetes
No full textIntra-day Variability of Glucose Absorption and Insulin Sensitivity: Assessment from AP@home Clinical Trial Data
No full textA Bayesian Method for the Identification of the Glucose-Insulin Model in Type 1 Diabetes
No full textCloning a Day of T1DM Individual Subjects from the FDA-accepted Simulator by a Bayesian Approach
No full textOne-Day Bayesian Cloning of Type 1 Diabetes Subjects: Toward a Single-Day UVA/Padova Type 1 Diabetes Simulator
No full textObjective: The UVA/Padova Type 1 Diabetes (T1DM) Simulator has been shown to be representative of a T1DM population observed in a clinical trial, but has not yet been identified on T1DM data. Moreover, the current version of the simulator is "single meal" while making it "single-day centric," i.e., by describing intraday variability, would be a step forward to create more realistic in silico scenarios. Here, we propose a Bayesian method for the identification of the model from plasma glucose and insulin concentrations only, by exploiting the prior model parameter distribution. Methods: The database consists of 47 T1DM subjects, who received dinner, breakfast, and lunch (respectively, 80, 50, and 60 CHO grams) in three 23-h occasions (one open-and one closed-loop). The model is identified using the Bayesian Maximum a Posteriori technique, where the prior parameter distribution is that of the simulator. Diurnal variability of glucose absorption and insulin sensitivity is allowed. Results: The model well describes glucose traces (coefficient of determination R-2 = 0.962 +/- 0.027) and the posterior parameter distribution is similar to that included in the simulator. Absorption parameters at breakfast are significantly different from those at lunch and dinner, reflecting more rapid dynamics of glucose absorption. Insulin sensitivity varies in each individual but without a specific pattern. Conclusion: The incorporation of glucose absorption and insulin sensitivity diurnal variability into the simulator makes it more realistic. Significance: The proposed method, applied to the increasing number of long-term artificial pancreas studies, will allow to describe week/month variability, thus further refining the simulato
Smart Algorithms for Efficient Insulin Therapy Initiation in Individuals With Type 2 Diabetes: An in Silico Study
No full textBackground: Insulin-naive subjects with type 2 diabetes (T2D) start basal insulin titration from a low initial insulin dose (IID), which is adjusted weekly or twice per week based on fasting plasma glucose (FPG) measurement as recommended by the American Diabetes Association (ADA). The procedure to reach the optimal insulin dose (OID) is time-consuming, especially in subjects with high insulin needs (HIN). The aim of this study is to provide a fast and effective, but still safe, insulin titration algorithm in insulin-naive T2D subjects with HIN. Method: To do that, we in silico cloned 300 subjects, matching a real population of insulin-naive T2D and used a logistic regression model to classify them as subjects with HIN or subjects with low insulin needs (LIN). Then, we applied to the subjects with HIN both a more aggressive insulin dose initiation (SMART-IID) and two newly developed titration algorithms (continuous glucose monitoring [CGM]-BASED and SMART-CGM-BASED) in which CGM was used to guide the decision-making process. Results: The new titration algorithm applied to HIN-classified individuals guaranteed a faster reaching of OID, with significant improvements in time in range (TIR) and reduction in time above range (TAR) in the first months of the trial, without any clinically significant increase in the risk of hypoglycemia. Conclusions: Smart basal insulin titration algorithms enable insulin-naive T2D individuals to achieve OID and improve their glycemic control faster than standard guidelines, without jeopardizing patient safety
The University of Virginia/Padova Type 1 Diabetes Simulator Matches the Glucose Traces of a Clinical Trial
No full text
Short- and Long-Term Effects on Glucose Control of Nonadherence to Insulin Therapy in People With Type 2 Diabetes An In Silico Study
No full textBackground: Strict adherence to multiple daily insulin (MDI) therapy is a cornerstone for the achievement of good glucose control in people with advanced type 2 diabetes (T2D). Here, we aim to in silico assess glucose control in T2D subjects with poor adherence to MDI therapy. Methods: We tuned the Padova T2D Simulator, originally describing early-stage T2D physiology, around advanced T2D people. One hundred in silico advanced T2D subjects were generated and equipped with optimal MDI therapy: specifically, basal and bolus insulin amounts and injection times were individualized for each subject by applying titration algorithms that iteratively update insulin dose based on glucose deviation from its target. Then, the effect of nonadhering to MDI therapy was assessed using standard glucose control metrics calculated in two 6-month 3-meal/day in silico scenarios: in Scenario 1, subjects received the optimal basal and prandial insulin bolus at each meal; in Scenario 2, subjects received optimal basal insulin and randomly delayed or skipped the prandial insulin bolus in 3 lunches during working days and 1 dinner during weekends. Results: A statistically significant degradation was found in all glucose control outcome metrics in Scenario 2 versus Scenario 1: e.g., percent time above 180 mg/dL increased by 22.2% and glucose management index by 0.2%. Conclusions: Impaired adherence to MDI therapy in T2D leads to glucose control deteriorations in both short and long terms. Interestingly, short-term hyperglycemia seems being contrasted by residual endogenous insulin secretion, which statistically increased by 3-fold after delayed/skipped insulin boluses compared with optimal ones
- …
