Final Report Summary - SOLARE (Seeking solutions for the artificial pancreas: new methods for improving continuous glucose monitoring and closed-loop postprandial glycaemic control)
Diabetes is chronic metabolic disorder that in 2012 affects 371 million people worldwide (8.3 % of the population between 20 - 79 years old) and 55 million people in Europe (6.7 %), according to the International Diabetes Federation. The global health expenditure related to diabetes amounted this year to USD 471 billion, of which 30 % corresponds to Europe. Approximately 10 % of the diabetic population is affected by type 1 diabetes, requiring external insulin administration for survival. Currently, a cure for type 1 diabetes does not exist, and the patient must rely on the technology to control blood glucose into normal levels.
The artificial pancreas consists in providing insulin pumps with enough intelligence to automatically deliver insulin based on continuous measurements of glucose (closed-loop glucose control). It is considered as the technological 'holy grail' for alleviating the burden of glucose control and the achievement of therapeutic goals in patients with type 1 diabetes, until a cure is found. Although great progress has been done in the past five years in the field, many challenges remain open until the diabetic society can benefit from this new technology. Current clinical trials demonstrate the efficiency of the artificial pancreas in reducing the incidence of hypoglycaemia and variability at night. However, performance after a meal is still poor with frequent induction of hypoglycaemia due to the controller overreaction.
SOLARE is a Marie Curie Intra-European Fellowship (IEF) project (Grant Agreement No 252085) addressing the challenge of an efficient and safe artificial pancreas for glucose control after a meal in patients with type 1 diabetes. The endocrinologist Dr Paolo Rossetti joined for two years the University Institute of Control Systems and Industrial Computing (ai2 Institute) at Universitat Politècnica de València, Spain, hosted by the scientist-in-charge Dr Jorge Bondia. The project focused onalgorithms for the optimisation of postprandial control and on accuracy improvement of continuous glucose monitoring (CGM), one of the bottlenecks in the artificial pancreas development.
Firstly, optimisation of postprandial control was tackled from both open-loop (current insulin pump therapy) and closed-loop (control algorithms for the artificial pancreas) perspective. Insulin dosing in insulin pumps still remains an empirical process highly dependent on the patients' and physicians' skills. In this project a clinical trial was conducted at the collaborating University Clinic Hospital of Valencia to validate an open-loop CGM-based method for the automatic individualisation of insulin pump therapy. Twelve subjects underwent four mixed-meal studies comparing traditional therapy with the new dosing algorithm. The subjects were also monitored at home to get a mathematical model describing his / her metabolism and characterising physiological variability. As result of this study, the feasibility of CGM for patient's characterisation and insulin dosing individualisation was demonstrated. However, superiority to traditional therapy was not demonstrated due to unexpected high variability in the study. A reduction of this variability can be expected from closed-loop strategies. Thus, effort was devoted to the study of new control engineering techniques aiming at the reduction of the risk of hypoglycaemia after meals. Reference-conditioning-based methods were successfully validated using a Food and Drug Administration (FDA)-accepted type 1 diabetes simulator, as compared to current algorithms used in the existing artificial pancreas prototypes. A clinical validation study was designed and the protocol prepared. The host group will conduct this study during 2013.
Secondly, accuracy of continuous glucose monitors was addressed. Lack of accuracy of current devices, especially in hypoglycaemia, has led to its approval only as adjunctive to standard glucometers, and it is currently a limitation in the artificial pancreas performance. In part, inaccuracies may be due to the fact that the devices estimate plasma glucose from interstitial glucose measurements. Thus, an analysis of the plasma-interstitial glucose relationship was carried out. A hyperinsulinemic-euglycaemic clamp study was implemented in the project to study the effect of insulin on this relationship. Fourteen patients underwent two clamp studies with low and high insulin concentration. Plasma glucose was measured frequently and interstitial glucose measured through a CGM device. No significant effect of insulin was found in the data analysis performed. Plasma-interstitial glucose relationship was modelled independently of insulin concentration through a local-models technique, able to represent non-linearities as a combination of linear local models. The technique, novel in the field, was integrated into a new calibration algorithm for CGM devices. A first preliminary validation conducted with background data of the group demonstrated its feasibility. Validation conducted with the combined set of data from both studies in this project reinforced the previous results. The existence of local dynamics was demonstrated, inferring a non-linearity in the hypoglycaemic range. As compared to the last-generation CGM device used in the studies (Medtronic Paradigm VeoTM), the error was reduced in 30 to 50 %, depending on the algorithm configuration, in the overall glucose range, and 43 % in hypoglycaemia. This error reduction is very significant and may have a big impact in CGM, not only in the context of the artificial pancreas but also in its clinical use for patient's monitoring.
In conclusion, the project objectives were successfully met significantly advancing in new safer postprandial control strategies and better continuous glucose monitors for the artificial pancreas. Once successfully validated and put into the market, it is expected that the artificial pancreas will have a huge impact in the quality of life of patients with type 1 diabetes, achieving a better glycaemic control translating into the development of less co-morbidities and the reduction of health expenditure of national health systems.
Contact details: Dr Jorge Bondia
ai2 Institute, Universitat Politècnica de València, Valencia, Spain
email: jbondia@isa.upv.es
Dr Paolo Rossetti
Hospital Francesc de Borja de Gandia, Gandia, Spain
email: prossetti73@gmail.com
The artificial pancreas consists in providing insulin pumps with enough intelligence to automatically deliver insulin based on continuous measurements of glucose (closed-loop glucose control). It is considered as the technological 'holy grail' for alleviating the burden of glucose control and the achievement of therapeutic goals in patients with type 1 diabetes, until a cure is found. Although great progress has been done in the past five years in the field, many challenges remain open until the diabetic society can benefit from this new technology. Current clinical trials demonstrate the efficiency of the artificial pancreas in reducing the incidence of hypoglycaemia and variability at night. However, performance after a meal is still poor with frequent induction of hypoglycaemia due to the controller overreaction.
SOLARE is a Marie Curie Intra-European Fellowship (IEF) project (Grant Agreement No 252085) addressing the challenge of an efficient and safe artificial pancreas for glucose control after a meal in patients with type 1 diabetes. The endocrinologist Dr Paolo Rossetti joined for two years the University Institute of Control Systems and Industrial Computing (ai2 Institute) at Universitat Politècnica de València, Spain, hosted by the scientist-in-charge Dr Jorge Bondia. The project focused onalgorithms for the optimisation of postprandial control and on accuracy improvement of continuous glucose monitoring (CGM), one of the bottlenecks in the artificial pancreas development.
Firstly, optimisation of postprandial control was tackled from both open-loop (current insulin pump therapy) and closed-loop (control algorithms for the artificial pancreas) perspective. Insulin dosing in insulin pumps still remains an empirical process highly dependent on the patients' and physicians' skills. In this project a clinical trial was conducted at the collaborating University Clinic Hospital of Valencia to validate an open-loop CGM-based method for the automatic individualisation of insulin pump therapy. Twelve subjects underwent four mixed-meal studies comparing traditional therapy with the new dosing algorithm. The subjects were also monitored at home to get a mathematical model describing his / her metabolism and characterising physiological variability. As result of this study, the feasibility of CGM for patient's characterisation and insulin dosing individualisation was demonstrated. However, superiority to traditional therapy was not demonstrated due to unexpected high variability in the study. A reduction of this variability can be expected from closed-loop strategies. Thus, effort was devoted to the study of new control engineering techniques aiming at the reduction of the risk of hypoglycaemia after meals. Reference-conditioning-based methods were successfully validated using a Food and Drug Administration (FDA)-accepted type 1 diabetes simulator, as compared to current algorithms used in the existing artificial pancreas prototypes. A clinical validation study was designed and the protocol prepared. The host group will conduct this study during 2013.
Secondly, accuracy of continuous glucose monitors was addressed. Lack of accuracy of current devices, especially in hypoglycaemia, has led to its approval only as adjunctive to standard glucometers, and it is currently a limitation in the artificial pancreas performance. In part, inaccuracies may be due to the fact that the devices estimate plasma glucose from interstitial glucose measurements. Thus, an analysis of the plasma-interstitial glucose relationship was carried out. A hyperinsulinemic-euglycaemic clamp study was implemented in the project to study the effect of insulin on this relationship. Fourteen patients underwent two clamp studies with low and high insulin concentration. Plasma glucose was measured frequently and interstitial glucose measured through a CGM device. No significant effect of insulin was found in the data analysis performed. Plasma-interstitial glucose relationship was modelled independently of insulin concentration through a local-models technique, able to represent non-linearities as a combination of linear local models. The technique, novel in the field, was integrated into a new calibration algorithm for CGM devices. A first preliminary validation conducted with background data of the group demonstrated its feasibility. Validation conducted with the combined set of data from both studies in this project reinforced the previous results. The existence of local dynamics was demonstrated, inferring a non-linearity in the hypoglycaemic range. As compared to the last-generation CGM device used in the studies (Medtronic Paradigm VeoTM), the error was reduced in 30 to 50 %, depending on the algorithm configuration, in the overall glucose range, and 43 % in hypoglycaemia. This error reduction is very significant and may have a big impact in CGM, not only in the context of the artificial pancreas but also in its clinical use for patient's monitoring.
In conclusion, the project objectives were successfully met significantly advancing in new safer postprandial control strategies and better continuous glucose monitors for the artificial pancreas. Once successfully validated and put into the market, it is expected that the artificial pancreas will have a huge impact in the quality of life of patients with type 1 diabetes, achieving a better glycaemic control translating into the development of less co-morbidities and the reduction of health expenditure of national health systems.
Contact details: Dr Jorge Bondia
ai2 Institute, Universitat Politècnica de València, Valencia, Spain
email: jbondia@isa.upv.es
Dr Paolo Rossetti
Hospital Francesc de Borja de Gandia, Gandia, Spain
email: prossetti73@gmail.com