A novel pulse-modulated closed-loop artificial pancreas based on intravenous administration of insulin and glucagon

Abstract

Current attempts to implement and apply automated control systems for the management of glucose homeostasis in individuals with diabetes are partly successful. In most semi and closed loop control systems to mimic the action of a normal pancreas in diabetes patients, insulin is administered subcutaneously. However, the parameters for insulin diffusion and transport time constants are relatively large and have wide individual variations. Therefore, deviation from a normal meal can result in suboptimal euglycemic control. Stable and reliable closed loop feedback control using continuous glucose monitoring under these conditions is difficult and needs regular interventions from the user. This article describes the translation of the endocrine physiology of a normal pancreas to an electronic equivalent. With this translation, the complex effects of a direct intravenous pulsatile method of insulin and glucagon administration can be simulated in accordance with physiological observations in a healthy subject and has been built with standard electronic components. This device is applicable for any individual and under any condition to automatically maintain the desired optimal euglycemic condition. The insulin and glucagon response control is presented in the same physiological pulsatile manner as is observed in healthy individuals.

Background

Glucose metabolism and control are fundamental processes for general energy management in the human body. Every individual has a specific set of physiological parameters determining glucose appearance after a meal and glucose metabolism control, with an adequate amount of insulin and glucagon secreted by the pancreas gland.

Today, it is the standard procedure for individuals with insufficient insulin production to manage the glucose concentration to the euglycemic range with subcutaneous insulin injections, which need regular monitoring of the glycemic state. Additionally, the risk of hypoglycemia has to be monitored and possibly counteracted by externally administered glucose or glucagon. This method of insulin administration and glucagon or glucose compensation is far from ideal compared with the natural way it operates in a healthy individual.

Insulin is an indispensable hormone that metabolizes glucose in various cells and tissue types. Glucose for the central nervous system is dependent on glucose transporters to pass through the blood–brain barrier instead of insulin metabolism¹.

The healthy pancreas secretes an adequate amount of insulin dependent on the amount and rate of glucose in such a way that after a meal, the glucose concentration returns within one hour to the normal equilibrium between 4.5 and 5.5 mmol/L.

Glucose derangements in humans, such as hyperglycemia and hypoglycemia, are normally prevented by homeostatic control mechanisms of the precise antagonism between pancreatic insulin and glucagon secretion. Under fasting conditions, the euglycemic state or basic glucose concentration G_C is maintained by balancing insulin and counteracting glucagon secretion impulses. Additional prandial insulin responses will occur when the ambient glucose concentration exceeds a specific basic threshold level, G_B, of approximately 4.5–5.5 mmol/L. The euglycemic stability of G_B is preserved by a counter regulatory feedback loop controlled by glucagon from pancreatic alpha cells to prevent hypoglycemia.

Mathematical modeling related to insulin‒glucose interactions has been published in recent decades, of which the minimal model is one of the most popular examples². Many variants have been investigated for in silico validation via simulations^3,4. The subjects of insulin‒glucose interactions, such as glucose- and/or insulin appearance and metabolism, are described with ordinary differential equations. However, a realistic translation of euglycemic control behavior is only possible when the feedback configuration is available as an open and thus uncontrolled feedback loop. Furthermore, one must consider the influence of the gastrointestinal tract (GIT) as an input source for the blood compartment to describe glucose appearance dynamics and related pancreas responses^5,6. A normal glucose control system is a closed loop that counteracts any external disturbance. Therefore, it is difficult to analyze and characterize such a system under these conditions.

Sophisticated control models from the universities of Cambridge, Harvard and Padova have been proposed, which were further developed for closed loop artificial pancreas applications^7–9. These models have at least 12 parameters, which primarily concern the effects of insulin, glucose disposal and control of glucose levels within the euglycemic range but neglect the gastrointestinal absorption process and the individual differences in the nonlinear mutual dependence of parameters.

Furthermore, the onset and delay properties of subcutaneously (SC) or intraperitoneally (IP) administered insulin must be taken into account for the feedback properties of an artificial pancreas solution⁵. However, the internal glucose and insulin dynamics cannot be observed in real time or directly followed in every organ of an individual.

Today, the system response to glucose input from a mixed meal can be monitored via the continuous measurement of blood glucose and C-peptide analysis for the monitoring of insulin dynamics. In accordance with publications, the use of the gastrointestinal tract as an input glucose source for the blood compartment clearly describes glucose appearance dynamics and related pancreas responses⁵.

In 2017, an article was published that describes the detailed patient-specific different gastrointestinal tract glucose responses on a mixed meal measured at 5-minute intervals, revealing that every individual is characterized by a specific glucose appearance rise time constant τ⁶. This resulted in the acceptance of a resistor capacitor model that characterizes the glucose appearance properties of an individual, which can reliably be used for in silico and real component simulations and emulations.

Recent developments have also introduced a semiautomatic bihormonal (subcutaneous insulin-glucagon) artificial pancreas (INREDA), which has been clinically evaluated¹⁰.

Glucose control physiology is a fully continuous analog system in which the insulin and glucagon responses to changes in the time of the pancreas with respect to changing glucose concentrations are characterized by time delays and transport time effects.

The transport delays, also known as dead time effects, inevitably result in control problems when an artificial pancreas linear feedback configuration operates in continuous time. Close observation of the pancreatic secretion of insulin and glucagon in healthy subjects revealed that specific pulsatile hormone secretion was dependent on the glucose concentration, with time intervals between 5 and 7 min¹¹.

This discrete pulsatile operation implies that evolution has revealed a way to achieve stable euglycemic glucose control on the basis of a sampling procedure of the pancreas to eliminate transport time and delay effects.

The article is organized as follows.

Section “Historical overview” presents a historical overview of currently published articles about automated glucose control based on subcutaneously administered insulin and their limitations.

Section “Artificial pancreas with subcutaneous insulin infusion” gives an overview of the currently available artificial pancreas solutions based on subcutaneously administered insulin for closed loop control applications.

Section “Materials and methods”, presents the translation procedure of the gastrointestinal tract, mixed meal consumption and related glucose appearance profile to an electrical network and network simulations with the use of the network simulator SIMetrix¹².

Section “Discrete time glucose control” compares the properties of single intravenously administered insulin bolus effects with those of multiple insulin bolus effects.

Section “Result” presents the complete closed loop control of an intravenously administered micro bolus insulin/glucagon system with simulated results covering the complete range of glucose control.

Section “Discussion” presents the discussion.

Historical overview

Artificial pancreas (AP) systems for automated homeostatic glucose control are considered the holy grail of diabetes management. The first attempts at glucose-responsive insulin delivery were pioneered in the 1960s and 1970s, with the.

development of systems that use venous glucose measurements to direct intravenous infusions of insulin and dextrose to maintain euglycemia^13–16.

Current advances in continuous glucose monitoring technologies and miniaturizing insulin and glucagon delivery systems have progressed to compact, wearable devices.

In addition to the use of open loop systems, there have also been attempts to introduce closed loop systems intended for fully automated homeostatic glucose control. These systems were implemented with the use of interstitial glucose sensing, subcutaneous insulin pumps, and increasingly sophisticated predictive control algorithms.

However, until recently, a fully operational substitute for normal pancreas operation in terms of individually detected glucose dynamics has not been identified, let al.one realized. Challenges in automated closed-loop technology include the development of systems that do not require manual intervention for meal announcements or carbohydrate counting. Another evolving avenue in research is the addition of glucagon to mitigate the risk of hypoglycemia, which allows for a wider margin of insulin administration¹⁰.

In the last 100 years, since the discovery of insulin, there have been technological advances in diabetes management. Insulin pumps became clinically feasible in the 1970s and were developed into more reliable miniaturized versions. Currently, continuous glucose monitoring systems (CGMS) are minimally invasive and relatively accurate. This opens new possibilities for designing closed loop glucose control systems that mimic the physiological functions of a healthy pancreas. The development of practically usable insulin–glucagon IV infusion methods is the next step in glucose control technology.

In recent years, there has been increased interest and research in the literature concerning closed-loop systems¹⁷. However, the desired level of automation cannot be achieved with the current approaches, as discussed above. The variety of glucose–insulin delays and dead time effects make it a challenge to design stable continuous-time analog feedback control, where the delay effects of subcutaneously administered insulin always result in unexpected feedback instabilities, and attempts to apply predictive control algorithms have achieved limited success¹⁴.

In Fig. 1, the glucose control block diagram of a normal pancreas is presented from a systems perspective. In a healthy individual, the homeostatic glucose concentration is maintained between 4.5 and 5.5 mmol/L. After the consumption of a meal, the glucose concentration increases over time, depending on the individual’s time constant τ, during which process the healthy pancreas responds with adequate amounts of insulin^5,6. After the return of the glucose concentration to the homeostatic value of approximately 5 mmol/L, a further decrease will be counteracted, with a normal pancreatic response of the α cells reducing β-cell activity and resulting in the secretion of glucagon. Both insulin and glucagon maintain a narrow glucose concentration range of approximately 5 mmol/L.

The glucose concentration control function is depicted as a block diagram in Fig. 1.

Fig. 1 [Images not available. See PDF.]

Block diagram of the glucose control function.

Figure 1 shows that the appearance of glucose is the result of the consumption of a mixed meal input via the gastrointestinal tract (GIT) to a summation point S with positive source inputs from the GIT, insulin and glucagon control. The glucagon source in this example is indicated as Rref of 5 mmol/L, and the same reference is valid for insulin control.

The glucagon source represents the α-cell action of the pancreas, which serves as a compensatory mechanism against a too low glucose concentration which prevents acute hypoglycemia. The normal glucose backup sources for energy metabolism are provided by glycogen from the liver and lipid tissues. This implies that a glucose level lower than 5 mmol/L results in a source activity from the glucagon control for the glucose concentration to maintain the glucose level at least 5 mmol/L. Insulin control is activated when the glucose level exceeds the reference level of 5 mmol/L.

The measured results from the blood compartment are subject to minimum phase delays and dead time transport effects before being processed for control actions in comparators 1 and/or 2. The time period between insulin or glucagon pulses is approximately 5 to 7 min, intended to wait long enough to decide what kind of action and with respect to what necessary amplitude a secretory response to the measured level of glucose has to be provided¹¹. Glucagon and insulin are both secreted in a pulsatile manner and are directly infused in the portal vane to the liver¹⁵.

Artificial pancreas with subcutaneous insulin infusion

Current implementations of semi-closed loop and fully closed loop artificial pancreas products are based on the subcutaneous injection of insulin by means of insulin pumps directed by a glucose controller and continuous measurement of the glucose concentration^7–10,16. The glucose–insulin–glucagon diagram of Fig. 1 presents a normally operating control with a glucagon and an insulin loop. The common configuration of commercially available artificial pancreas products is presented in Fig. 2.

Fig. 2 [Images not available. See PDF.]

Conventional artificial pancreas configuration with single loop subcutaneous insulin glucose control.

A closed loop pancreas system with subcutaneously administered insulin encounters a variety of transport time effects, delays and nonlinearities demanding a regular user intervention when the system indicates for the necessary attention.

This approach has extensively been modeled^2,7–9.

Materials and methods

The translation of mixed meal consumption to an electrical equivalent

From an earlier published article describing the translation of mixed meal consumption to the glucose appearance profile of diabetes type 1 patients during a trial under clinically controlled conditions, the following is representative of all individuals examined in this trial⁶.

In Fig. 3, a representative example of the glucose profile of patient P23 from this trial is presented. This example is similar to other analyzed cases from the same trial.

Fig. 3 [Images not available. See PDF.]

The glucose profile of P23 according to the original trial dataset. The right panel shows the electrical translation of this profile.

With the modeling method described in this article, the physiological glucose response to a mixed meal can be translated to the electrical model shown on the right side of Fig. 3.

The capacitor model for the glucose compartment is therefore validated to be acceptable as a reliable model for further simulations in our newly proposed artificial pancreas.

According to the glucose charging profile of Fig. 3 in the left panel, which is represented by the dashed straight lines, the same charge profile is used for the voltage source U₁ in the right panel of Fig. 3. The meal ended after 15 min and reached a hold level of 14.2 mmol/L until t = 100 min, as shown in Fig. 3. With the characterized time constant of the glucose appearance of P23 indicated in the left panel (τ = 2000 on the second scale), the value of R₁ can be calculated according to R₁C₁ = τ, with a known value of C₁ = 0.3 F combine with a value of R₁ = 6600 Ω⁶.

The maximum value of the appearance profile is reached at t = 100 min = 6000 s.

The electrical behavior of the profile in Fig. 3, including the electrical configuration of the glucose decay behavior, can be visualized with the use of a suitable simulator, SIMetrix, which renders the results depicted in Fig. 4.

This profile similarity has been emulated and verified with a simulation as depicted in Fig. 4. The subcutaneously administered bolus insulin is based on the experiences of the patient’s response on a predefined amount of expected glucose after standard test meal.

Fig. 4 [Images not available. See PDF.]

Emulation (circuit on the upper panel) and simulation of the glucose appearance (green curve) after a standard test meal using a predefined subcutaneously administered insulin bolus (red curve).

In Fig. 4 the upper panel presents an extension of the circuit in Fig. 3 with a constant current discharge transistor Q₁. Voltage source U₂ represents the effect of the subcutaneously injected insulin of 18 units for P23, activating the resistor capacitor combination R₄, C_3, representing the constant active time of insulin in the blood stream for 133 min. Resistor R₃ represents the individual resistance of the insulin effect on the decay of U_C1, representing the glucose concentration. In the lower panel, the glucose appearance (U_C1) and decay are shown as green lines. The subcutaneous insulin profile U2 is indicated in red as a function of time.

Effect of intravenously administered insulin

With the circuit configuration shown in Fig. 5, the glucose level can also be decreased with an intravenously administered single insulin bolus or with a series of smaller bolus pulses generated by source U₂. The results of the glucose appearance and the decay effects of the insulin pulses are depicted in the lower panel of Fig. 5.

Fig. 5 [Images not available. See PDF.]

The upper panel shows the effect of a single insulin bolus (red) with variation in the parameter R₃, which represents the individual insulin resistance of the given bolus. In the lower panel, the effect of a series of 12 intravenously administered small insulin bolus pulses with variation in the physilological parameter R₃ is shown.

The presented effect of the series of insulin bolus pulses in the lower panel of Fig. 5 reveals a similar glucose decay effect, as could be observed from the original P23 appearance and decay profile in Fig. 3. This result implies that the micro bolus pulse series can effectively be deployed in a discrete architecture for glucose control.

Discrete time glucose control

With the simulation results presented in Fig. 5, it is feasible to design a control architecture based on repeated insulin and glucagon pulses instead of a single bolus to achieve the same euglycemic control results as a normal pancreas.

Several studies have analyzed and presented the effects of pulsatile secreted insulin and glucagon^11,18,19.

In addition to published observations and modeling of the pulsatile administration of insulin and glucagon, a detailed analysis and simulation of this subject in the form of an artificial pancreas function has not been published^19,20.

Mirbolooki et al.¹⁹ reported an important review about best practices for the treatment of diabetes mellitus types 1 and 2.

The pulsatile insulin administration approach with small boluses at 6-minute intervals under clinically controlled conditions revealed a remarkable improvement compared with a subcutaneously administered single bolus insulin¹⁹. Although the underlying mechanism is not understood, this principle forms the basis of our proposed artificial pancreas approach.

The results of Farhy et al. inspired the design and implementation of a dual-hormone pulse-modulated artificial pancreas^21,22. In particular, the Figs. 5, 6 and 7 of Farhy et al.²² present the signal template for the new design.

Using the architecture depicted in Fig. 6, the glucose appearance from the gastrointestinal tract (GIT) is measured in the blood stream after the summation point S, which is now based on the effects of intravenously administered insulin and/or glucagon.

Fig. 6 [Images not available. See PDF.]

Functional block diagram of the pulse-controlled artificial pancreas.

The diagram in Fig. 6 shows an endocrine translation of normal operating pancreas function, as depicted in Fig. 1, with the addition of a sample and hold function, controlled by a suitable timing circuit, to monitor the blood glucose level in both the glucagon and insulin control branches. This architecture results in a stable glucose control design where the disturbing effects of transport (dead time) delays are eliminated by the sample and hold function, and a fully closed loop operation can be realized¹⁷.

In the design of pulse-controlled dual loop artificial pancreas (AP), the properties and specifications of the insulin/glucagon pump are highly important.

As discussed in the architecture of the AP in Fig. 6, the pump operation is pulse controlled.

The injected insulin/glucagon micro bolus is dependent on the actual sampled G_C value, which determines the pump pulse amplitude and, as such, the added new amount of insulin/glucagon.

Because insulin/glucagon has a half-life of approximately 5–6 min in plasma, a remaining amount of the injected hormones will be added to the next administered dose. This phenomenon is known as pharmacokinetic accumulation, the effect of which is illustrated in the bottom panel of Fig. 7.

With a thought experiment, the new artificial pancreas is assumed to operate as follows.

The G_C is measured every 5 min on the basis of the insulin and glucagon half-life of 5 min. When the G_C concentration exceeds 5 mmol/L, a small insulin bolus (too small for an overdose) proportional to the G_C concentration is intravenously injected. After 5 min, a new sample of the G_C was measured. The new value of G_C > 5 mmol/L is a proportional measure for the insulin pump to administer a following insulin bolus. The administered insulin bolus is added to the remaining insulin in the blood stream. This process continues until the G_C no longer increases. A decaying G_C, measured after 5 min, will result in a reduced insulin bolus proportional to the measured G_C value. This method does not require a model but acts on emergent intermittent findings after a waiting time of 5 min.

Results

A functional block diagram of the described sampling operation and the insulin/glucagon pump control together with the simulation results is presented in Fig. 7. According to this natural principle, the pulse height is modulated proportionally to the measured G_C appearance and added to a pharmacokinetic impulse shaper, which functions to reduce the glucose level. The glucose level will therefore be sensed with sampling periods of 5 min, according to the insulin half-life in plasma.

Fig. 7 [Images not available. See PDF.]

At the top, a block diagram and schematic of the complete artificial pancreas are presented. The electronic circuit presented in the middle panel of the figure is intended for simulation purposes to validate and verify the proposed system in real time operation¹². At the bottom, the simulation result is presented as the response of a rising glucose concentration (green line) from zero to the maximum of Gc. The control pulses and their pharmacokinetic effect are indicated as glucagon pulse (blue)/insulin pulse (red). In this example, the simulation reveals the effect of generated glucagon/insulin pulses on glucose metabolism and the final euglycemic control result of 5 mmol/L.

The electronic schematic diagram of Fig. 7 shows that the linear voltage-to-current conversion ΔI(Gc) = U_Gc/R₆ is comparable with the 5 mmol/L reference V3 for the G_C and is processed with transistors Q11 and Q12. The linear signal for insulin control (G_C > 5 mmol/L) is transferred via current mirrors Q25 and Q13 to insulin pulse modulators Q17, Q18 and Q19 to insulin blood compartments R30, C1, and Q20, respectively, to discharge the glucose compartment capacitor C3. Similarly, the glucagon control arm is activated when the G_C < 5 mmol/L, and the linear proportional glucagon signal is transferred as a glucagon signal current via the current mirror Q23 and Q14 to the glucagon pulse modulators Q10, Q6, and Q7 and transferred via the current mirrors Q24 and Q8 to the glucagon compartments R32, Q4, and Q1 to activate the addition of glucagon pulses to the blood compartment capacitor C3 via the current mirrors Q2 and Q3. The G_C as a voltage on C3 will be buffered by the series of complementary emitter followers Q4, Q22 to mimic the continuous glucose monitor function and provide this voltage as a feedback signal to the base of Q11 in the linear comparator Q11, Q12.

The timing pulses of V5 are processed by the modulator switches Q6 and Q7 for the glucagon pulse switch and Q18 and Q19 for the insulin pulse switch.

Discussion

In this work, the glucose control system analysis is based on individual observations.

The modeling results of postprandial glucose appearance have been used as the resistor capacitor glucose compartment in simulations of our proposed artificial pancreas⁶.

Subcutaneously administered insulin is still the common method used by insulin-dependent individuals and is sometimes supported by means of semiautomatic or semi closed loop control insulin pumps relying on subcutaneous insulin administration. This method of single (estimated) bolus insulin administration has a relatively long half-life⁵ in the tissue, which ensures a rather constant metabolic effect over several hours, which was measured as a constant current discharge characteristic, as presented in Fig. 3. On the other hand, the relatively large transport time, with a strict individual value, also represents an uncontrollable transport time phase shift for the insulin pump controller, which complicates a stable control result. Schiavon et al.⁵ characterized both the subcutaneous (SC) and the intraperitoneal (IP) routes of insulin administration. Figure 3 in this article⁵ shows that the onset time (time before reaching Cmax) varies from approximately 30 min for IP to 60 min for SC. The estimated half-life from the same figure is then one hour for IP and 2 h for SC.

This finding implies that, regardless of the method of administration, pulsatile^19,24–26 or subcutaneous, the insulin response to a necessary control signal from the continuous glucose monitor takes at least 30 and 90 min to cause a measurable effect from the insulin pump infusion detected by the glucose monitor. Because of this insulin transport effect, stable closed loop control is difficult to achieve. Nevertheless, this method has the ability to manage the necessary glycemic state for patients with type 1 diabetes, although the user needs to intervene regularly to maintain the glucose concentration within a safe range. This control action delay can be described as follows.

When the glucose concentration change is detected by the glucose sensor, the control action sent to the SC or IP insulin pump has a greater effect after approximately one hour. The situation can be compared with that of a car driver who cannot see the road while driving in a car at 70 km/h. The driver is guided by an observer to provide instructions for a safe passage along the road. When the observer indicates a necessary turn to the right, the message to take action for the driver arrives too late, and the car cannot make the necessary turn in time, resulting in an off-road position. This happens with fully closed loop glucose monitoring combined with an SC or IP insulin pump. Users with such a treatment device for the management of diabetes type 1 need to intervene at regular moments of the day to keep the system on the road and, in all cases, avoid the state of hypoglycemia. Recently, pulsatile subcutaneous insulin administration methods have also been proposed, and simulation results have indicated an improvement in glycemic control^24–26.

However, although the pulsatile designed administration of subcutaneous insulin described by Cocha et al.²⁵ was reported to improve glycemic control, this method still does not solve the feedback delay problem, and the user must intervene regularly for corrections.

In this article, we present validated translations of glucose physiology to electrical and electronic networks. A similar publication used this method for detailed modeling at the cell level²³. Without the complexity of all related and coupled differential equations, the changes in stimuli of the glucose control system can easily be examined and evaluated with the use of electronic networks and an adequate network simulator to acquire an intuitive look and feel for the resulting responses.

This method opens a wide variety of design possibilities and simulation scenarios for an artificial pancreas that operates in real time. With our approach, the natural pancreas operation is mimicked by means of a pulse-modulated glucose–insulin–glucagon control model implemented with electronic circuits. When such a circuit fulfills the control requirements, artificial pancreas function can be realized with an intravenously injected micro bolus of insulin or glucagon, which is dependent on the actual glucose concentration.

To achieve this type of glycemic control, it is necessary to administer insulin and/or glucagon intravenously. This administration method is currently not used in the treatment methods and guidelines for diabetes patients but is a known and mature technology in the treatment of cancer patients who need regular intravenous administration of their cancer medication^27,28.

The paradigm of continuous feedback control for glucose homeostasis has to be changed to the method of time discrete system configurations to avoid control problems caused by lag and dead time effects. The diffusion and transport effects of subcutaneously administered insulin can be replaced by a controlled intravenous micro bolus, which is small enough to prevent harmful overdose because there is sufficient time and space for the necessary corrections. Here, the introduction of an electronic circuit implementation of such a system realizes the desired features for optimal control by sampling the actual glucose concentration at specific moments.

According to the simulation results, the adaptive properties of the controller imply that a learning period is possibly very short because all glucose-dependent scenarios are automatically adapted to individual physiology with an adequate insulin/glucagon micro bolus. In this way, intravenously administered insulin/glucagon can act as a reasonable approximation of a normal healthy pancreas operation.

The presented concept of an artificial pancreas presents an alternative for the traditional approach of subcutaneously administered insulin and offers various possibilities to simulate the glucose control system. Both students and clinicians can design all kinds of scenarios to test different individual physiological variables, such as the following:

1. Values of the homeostatic glucose concentrations.

2. Variations in the sampling time.

3. Variations in pulse amplitude conformity.

4. Variations in pulse width.

5. Variations in insulin and glucagon half-life parameters.

6. Variations in the individual insulin resistance values.

For future studies and development with real patients, the following is of special interest.

With this simulation of a healthy pancreas, the insulin/glucagon administration method is a point of concern. The insulin/glucagon impulses are supposed to be administered intravenously, resulting in a fast physiological response and a relatively short insulin glucose interaction time.

This administration regimen needs further study and development in close cooperation with experienced diabetologists and mechanical engineering specialists to construct a fail-safe insulin–glucagon administration method in a clinically controlled environment and follow-up with the necessary clinical trials^27,28.

The proposed artificial pancreas can be tested under clinical conditions for the critical treatment methods necessary in cases of acute hyperglycemia.

Finally, this design is expected to provide individuals with diabetes mellitus type 1 with the desired solution for unattended and safe operation of the artificial pancreas for optimal glycemic control regardless of the meal to be consumed and/or physical exercise.

Thus, it is to be expected that these people can have a better quality of life and possibly a longer life span.

Nevertheless, the possibility of managing the glycemic state of people with type 1 diabetes via the combination of subcutaneous administration with an insulin pump and glucose monitoring is regarded as a major achievement in the currently available treatment methods.

Author contributions

S.L.G. mainly wrote the manuscript. V.H.S., U.W. and W.J.H.B. were primarily responsible for study supervision and scientific evaluation. J.P.C.B. was associated as professor (lector). All authors reviewed the manuscript.

Data availability

The datasets used and/or analysed during the current study available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Steen, G. et al. No effect of insulin on glucose blood–brain barrier transport and cerebral metabolism in humans. Diabetes, Vol. 48, October (1999).

2. Bergman, RN; Ider, YZ; Bowden, CR; Cobelli, C. Quantitative estimation of insulin sensitivity. Am. J. Physiol.; 1979; 236, pp. E667-E677.[COI: 1:CAS:528:DyaE1MXksFelsL0%3D] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/443421]

3. Palumbo, P., Ditlevsen, S., Bertuzzi, A. & De Gaetano, A. Mathematical modeling of the glucose-insulin system: A review. Math. Biosci.244 (69–81), (2013).

4. Chee, F. & Fernando, T. Lecture notes in control and information sciences, close-loop control of blood glucose. Springer Verlag ISBN: 978-3-540-74030-8 (Print) 978-3-540-74031-5 (2007).

5. Schiavon, M., Cobelli, C. & Dalla Man, C. Modeling intraperitoneal insulin absorption in patients with type 1 diabetes. Metabolites11, 600. (2021). https://doi.org/10.3390/metabo11090600 (2021).

6. Goede, S. L., de Galan, B. E. & Leow, M. K. Personalized glucose- insulin model based on signal analysis. J. Theoret. Biol.419 333–342 (2017). https://doi.org/10.1016/j.jtbi.2016.12.018 (2017).

7. Daly, A. B. et al. Fully automated closed-loop insulin delivery in adults with type 2 diabetes: An open-label, single-centre randomised crossover trial. Nat Med 11 Jan; https://doi.org/10.1016/j.jtbi.2016.12.01810.1038/s41591-022-02144-z (2023) (2023).

8. Toffanin, C., Magni, L. & Cobelli, C. Artificial pancreas: In silico study shows no need of meal announcement and improved time in range of glucose with intraperitoneal vs subcutaneous insulin delivery (IEEE transactions on medical robotics and bionics, 2020).

9. Dassau, E. et al. Twelve-Week 24/7 ambulatory artificial pancreas with weekly adaptation of insulin delivery settings: Effect on hemoglobin A 1c and hypoglycemia, Diabetes Care Volume 40, December (2017).

10. Blauw, H., Onvlee, A. J., Klaassen, M., van Bon, A. C. & DeVries, J. H. Fully closed loop glucose control with a bihormonal artificial pancreas in adults with type 1 diabetes: An outpatient, randomized, crossover trial. Diabetes Care; 44 836–838 https://doi.org/10.2337/dc20-2106 (2021).

11. Farhy, LS; McCall, AL. Optimizing reduction in basal hyperglucagon aemito repair defective glucagon counter regulation in insulin deficiency diabetes. Obes. Metab. Oct.; 2011; 13, 0 1 pp. 133-143.[COI: 1:CAS:528:DC%2BC3MXhtlaqsrjK] [DOI: https://dx.doi.org/10.1111/j.1463-1326.2011.01455.x]

12. SIMetrix, T. & Ltd Circuit simulation, 78 Chapel Street, Thatcham, Berkshire, RG18 4QN United Kingdom, (2017). http:/www.SIMetrix.co.uk

13. Jacquot, R. G. Modern digital control systems. Electrical and computer engineering. 2nd Edition, CRC Press Website http://www.crcpress.com/product/isbn/9780824789145 (1994).

14. Kelly, J. L. Continuous insulin infusion: When, where and how? Diabetes spectrum Volume 27, Number 3, (2014).

15. Vargas, F. L., Garelli, F., De Battista, H. & Vehí, J. Postprandial response improvement via safety layer in closed-loop blood glucose controllers. (2015). https://doi.org/10.1016/j.bspc.2014.10.003.

16. Moon, S. J., Jung, I. & Park, C. Y. Current advances of artificial pancreas systems: A comprehensive review of clinical evidence. 77 Diabetes Metab J; 45:813–839 pISSN 2233–6079 eISSN 2233–6087 (2021). https://doi.org/10.4093/dmj.2021.0177 (2021).

17. Aström, K., Murray, R. M. & Feed back systems. Princeton University Press, Princeton (2014).

18. Pörksen, N. et al. Pulsatile insulin secretion: Detection, regulation, and role in diabetes. Diabetes, VOL. 51, Supplement 1, February (2002).

19. Mirbolooki, M. R. et al. J.R.T. Pulsatile intravenous insulin therapy: The best practice to reverse diabetes complications? Medical Hypotheses Volume 73, Issue 3, September (2009).

20. Huang, M; Li, J; Song, X; Guo, H. Modeling impulsive injections of insulin: Toward artificial pancreatitis. SIAM J. Appl. Math. Vol; 2012; 72, 5 pp. 1524-1548. [DOI: https://dx.doi.org/10.1137/110860306]

21. Farhy, LS et al. Amplification of pulsatile glucagon counter regulation by switch-off of Į-cell-suppressing signals in streptozotocin-treated rats. Am. J. Physiol. Endocrinol. Metab. 2008 September; 2008; 295, 3 pp. E575-E585.[COI: 1:CAS:528:DC%2BD1cXhtFCgt77E] [DOI: https://dx.doi.org/10.1152/ajpendo.90372]

22. Farhy, L. S. & McCall, A. L. A pancreatic network control of glucagon secretion and counter regulation, Meth. Enzymol.467 547–581.https://doi.org/10.1016/S0076-6879(09)67021-X (2009).

23. Teo, J. J. Y. & Sarpeshkar, R. The merging of biological and electronic circuits, iScience23 101688 https://doi.org/10.1016/j.isci (2020).

24. Abuin, P., Rivadeneira, P. S., Ferramosca, A. & González, A. H. Artificial pancreas under stable pulsatile MPC: Improving the closed-loop performance, J. Process Control44 (10):756–764 https://doi.org/10.1177/03913988211027176 (2020).

25. Cocha, G; Tedesco, V; D’Attelis, C; Amorena, C. Artificial Pancreas under stable pulsatile behavior improves artificial pancreas performance. Int. J. Artif. Organs; 2021; [DOI: https://dx.doi.org/10.1177/03913988211027176journals.sagepub.com/home/jao] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34348505]

26. Garcia-Tirado, J. et al. Assessment of meal anticipation for improving fully automated insulin delivery in adults with type 1 diabetes. Diabetes Care 1 September (2023).

27. Glatot, F. et al. February. Randomized phase 2 trial evaluating the safety of peripherally inserted catheters vs implanted port catheters during adjuvant chemotherapy in early breast cancer patients. The Eur. J. Cancer126 116–124 (2020).

28. Lee, J., Hur, S. H., Kim, Z. & Lim, C. W. Safety of immediate use of totally implantable venous access ports in adult patients with cancer: a retrospective single-center study. Korean J. Clin. Oncol. (2021).