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

María F Villa-Tamayo; Patricio Colmegna; Marc D Breton

doi:10.1177/19322968231212205

. 2023 Nov 15;18(2):318–323. doi: 10.1177/19322968231212205

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

María F Villa-Tamayo ¹, Patricio Colmegna ¹, Marc D Breton ^1,^✉

PMCID: PMC10973857 PMID: 37966051

Abstract

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.

Keywords: rebound hypoglycemia, automated insulin delivery systems, hypoglycemia, type 1 diabetes, safety layer

Introduction

Automated insulin delivery (AID) systems represent a game-changing approach to treating type 1 diabetes (T1D).^1-3 These advanced closed-loop systems aim to maintain the blood glucose (BG) level in a healthy range by integrating continuous glucose monitoring (CGM) sensors with insulin pumps through control algorithms that automatically regulate insulin infusion.³ AID systems can operate in a hybrid closed loop (HCL) mode, requiring manual announcements, or a full closed loop (FCL) mode, which may not require user interaction.¹ Currently, commercially available AID systems are labeled for HCL use, with ongoing research into the FCL mode.^1,2,4 The Inreda system has become the only FCL AID system approved by the European Commission.^3,5

While AID systems offer significant benefits in improving glycemic control and reducing the burden of constant diabetes management, they are not exempt from limitations.⁶ One crucial consideration is the risk of hypoglycemia, a condition that develops as an interplay of insulin excess and the compromised glucose counter regulation in T1D⁷ and that is linked to consequences such as dizziness, blurred vision, cognitive impairments, and in extreme cases, seizures and diabetic coma.^6,8-10

Different strategies have been devised to prevent hypoglycemia episodes in diabetes management. Monitoring systems now incorporate low-threshold alerts and/or hypoglycemia prediction alarms to enable users to take proactive measures.^11,12 Personalized hypoglycemia predictive alerts have also been introduced using glucose-insulin models or long short-term memory models to predict glucose trends.^9,13 In addition, different algorithms have been developed and integrated into AID systems to proactively reduce or suspend insulin infusion when hypoglycemia is predicted or present.^8,10,14-19

Despite these efforts, hypoglycemia events can still occur. When they do, people with T1D usually take fast-acting (or rescue) carbohydrates to rapidly restore normoglycemia.^20,21 However, this sudden change in BG levels can trigger an excessive control response, potentially inducing a second hypoglycemia episode. Lee et al²² reported this oscillatory rebound hypoglycemia during a clinical validation of their HCL controller, where the controller overreacted to rapid BG increases after rescue carbohydrates. To address this issue, a dynamic insulin-on-board constraint was proposed to minimize the likelihood of controller-induced hypoglycemia following a rapid rise of glucose levels.²² In addition, recent data from a clinical trial⁴ (ClinicalTrials.gov ID NCT04877730) evaluating HCL and FCL modes of the UVA AID system revealed instances of rebound hypoglycemia (see the Supplemental Material for an example of rebound hypoglycemia observed in real-life data). Specifically, there were five rebound events during the HCL mode and three in the FCL mode across the 35 subjects, and 9% of hypoglycemic episodes were attributed to rebound lows induced by controller overcorrection. This underscores the need for strategies to counteract the controller’s overreaction to rescue carbohydrates.

In this work, a controller agnostic HypoSafe module is developed to prevent rebound hypoglycemia. Based only on preceding glucose conditions, HypoSafe sets constraints on subsequent control actions without relying on manual user inputs. To evaluate the efficacy of this strategy, we integrated HypoSafe into the University of Virginia (UVA) AID system²³ within the Food and Drug Administration (FDA)-accepted UVA/Padova T1D simulator.²⁴ Rebound hypoglycemia events and glucose outcome metrics are reported for two different simulation scenarios: one including unannounced meals and another including an over-sized manual bolus.

Methods

HypoSafe Module

The primary goal of the HypoSafe module is to prevent rebound hypoglycemia by informing the system with conditions preceding the current glucose value and then determining when to allow full control actions. To that end, a dynamic upper bound over the control action is imposed. It is to consider both basal and bolus actions, especially in the case of FCL controllers that can react to BG rises by triggering a bolus.

Based on the analysis of a prior glucose trace, the following design principles are desired: (1) the lower the glucose levels have been in the recent past, the stricter the constraint should be; (2) the lower the current glucose value, the stricter the constraint; (3) if minimum glucose in the recent past is at or above the glucose target, then the constraint is maximum, allowing for full control actions; and (4) if hypoglycemia is detected in the recent past, there should not be any insulin delivery above the subject’s predefined basal rate profile $(B R_{p r f})$ until reaching the glucose target. Hypoglycemia was defined as BG levels under 70 mg/dL based on the American Diabetes Association guidelines.²⁵

Following these principles, the safety layer is then designed to apply an exponential decay constraint over the insulin delivery dose as a response to the minimum glucose value recorded in the last hour $(G_{m i n})$ and the current glucose concentration $(G_{c}) .$ The lower the values of $G_{m i n}$ and $G_{c},$ the stricter the constraint is meant to be. Considering a glucose target of 120 mg/dL, the insulin constraint is given by:

c o n s t r a i n t = 1.1 * m a x v a l u e * e^{- \frac{\max (0, a (120 - G_{m i n}))}{\max (0.1, b G_{c} - c)}} - 1

(1)

with $m a x v a l u e$ corresponding to the maximum desired value for the insulin signal that is being constrained, and factors $a = 0.14299,$ $b = 0.0714,$ and $c = 5.7143$ determined via numerical optimization to fit the curve that satisfy the design principles mentioned earlier.

HypoSafe can be applied to both basal and bolus commands with independent boundary constraints. For instance, in Figure 1, the set of curves depending on $G_{m i n}$ and $G_{c}$ can be observed for both basal (left axis) and bolus (right axis), where the ranges for each are a design selection depending on the signal to be modulated.

Integration With a Fully Closed-Loop Algorithm

The UVA AID system integrates (1) a safety system for imminent hypoglycemia mitigation (legacy SSM), (2) a hyperglycemia mitigation system (HMS), (3) a bolus priming system (BPS) to compensate for abrupt positive disturbances, and (4) a model predictive control (MPC) algorithm for background insulin regulation.

In summary, the MPC relies on a mathematical description of the insulin-glucose interaction to calculate an optimal control policy of insulin deviations from the subject’s predefined $B R_{p r f}$ that satisfies a set of constraints within a time horizon. As for the BPS, it estimates the probability of large positive disturbances (like meals) based on past glucose measurements. If the probability is above a certain threshold, the BPS triggers a bolus that depends on different ratios of the subject’s total daily insulin (TDI). For more details of both modules, please refer to Garcia-Tirado et al.²³

To prevent rebound hypoglycemia from basal and bolus commands, HypoSafe is coupled to both the MPC and BPS as depicted in Figure 2. For basal, the constraint was set to allow a maximum of 10 times the subject-specific $B R_{p r f}$ (U/min), ie, $m a x v a l u e = 10$ in (1), obtaining:

H y p o S a f e_{b a s a l} = B R_{p r f} * c o n s t r a i n t .

(2)

with $u_{b a s a l} = \min (u_{M P C}, H y p o S a f e_{b a s a l}) .$

In case of the BPS, the constraint was defined to represent fractions of the subject’s TDI (U) allowing up to 10% of its value $(m a x v a l u e = 10) :$

H y p o S a f e_{b o l u s} = \frac{T D I}{100} * c o n s t r a i n t .

(3)

The final priming bolus is given by $u_{b o l u s} = \min (u_{B P S}, H y p o S a f e_{b o l u s}) .$

Simulation Scenario and Outcomes

Experiments were run on the 100-adult cohort of the FDA-accepted UVA/Padova T1D simulator to assess the efficacy of the HypoSafe module. Two different scenarios were considered: (1) a nominal scenario consisting of 30 hours with three unannounced meals provided to each subject at 7 hours, 13 hours, and 19 hours with carbohydrate content of 70, 50, and 80 g, respectively, and (2) a 16-hour hypoglycemia-induced scenario where an overestimated manual bolus is provided for a single meal at 6 hours with 70 g of carbohydrates. For both scenarios, intra-day variability in insulin and dawn phenomenon is included, and hypoglycemia treatments of 15 g of fast-absorbing carbohydrates are provided every 20 minutes until glucose is above 60 mg/dL.

The primary outcome is the number of rebound hypoglycemia episodes caused by insulin delivery in the next 30 minutes of the first hypoglycemia event. A rebound hypoglycemia event was defined if BG <70 mg/dL within a two-hour window from a previous hypoglycemia event. In addition, basal and bolus commands within 30 minutes after a hypoglycemia event are reported. Overall glucose outcome metrics are also analyzed: mean BG, coefficient of variation (CV), and percentage of time spent between 70 and 180 mg/dL (time in range, TIR), above 180 mg/dL (time above range, TAR), and below 70 mg/dL (time below range, TBR), along the percentage of time in below 54 mg/dL, in a tight range of 70 to 140 mg/dL, and above 250 mg/dL.

Results

Figure 3 shows the evolution of glycemia, basal, and bolus of a representative subject under the closed-loop system with (pink) and without (blue) the HypoSafe module in the hypoglycemia-induced scenario. It is observed that around 11:30 am, a hypoglycemia event occurs given the overestimation of the meal bolus at breakfast. After the event, rescue carbohydrates are provided, rapidly raising the BG level. Without HypoSafe, basal infusion is increased, and a priming bolus is triggered, causing a second hypoglycemia event. It is also observed that the rebound hypoglycemia is repeated generating an oscillatory response. In contrast, when integrating the HypoSafe module into the AID system, both basal and bolus commands are safely constrained, avoiding in this way the rebound toward hypoglycemia. A detailed explanation on the computation of HypoSafe constraint for this example can be found in the Supplemental Material.

Population results for both scenarios are reported in Table 1.

Table 1.

Population Outcome Metrics in Nominal and Hypoglycemia-Induced Scenario.

	Nominal scenario		Hypoglycemia-induced scenario
outcomes	UVA FCL AID system	UVA FCL AID system+ HypoSafe	UVA AID system	UVA AID system + HypoSafe
Total rebound hypoglycemia events caused by insulin delivery	39	0	55	7
Basal delivery within 30 minutes after hypo event	0.27 ± 0.31	0.05 ± 0.09	0.27 ± 0.32	0.03 ± 0.04
BPS within 30 minutes after hypo event	1.02 ± 2.10	0.05 ± 0.19	0.43 ± 1.17	0.02 ± 0.08
Mean BG (mg/dL)	139.7 ± 9.68	141.9 ± 9.9	126.4 ± 12.9	128.1 ± 12.2
Coefficient of variation (%)	29.4 ± 6.4	29.0 ± 6.0	26.5 ± 7.5	25.6 ± 7.6
BG <54 mg/dL	0.2 ± 0.5	0.2 ± 0.3^a	0.2 ± 0.5	0.2 ± 0.5^a
BG <70 mg/dL	2.0 ± 1.6	1.7 ± 1.4	3.1 ± 2.1	2.7 ± 1.8
BG 70-140 mg/dL	61.3 ± 10.4	59.5 ± 10.7	75.6 ± 13.1	75.6 ± 12.9^a
BG 70-180 mg/dL	80.4 ± 8.6	79.8 ± 8.6	87.7 ± 9.2	88.1 ± 9.1
BG >180 mg/dL	17.6 ± 8.2	18.5 ± 8.3	9.2 ± 9.2	9.2 ± 9.3^a
BG >250 mg/dL	2.5 ± 4.1	2.7 ± 4.4	1.7 ± 4.4	1.7 ± 4.4^a

Open in a new tab

Abbreviations: UVA, University of Virginia; FCL, full closed loop; AID, automated insulin delivery; BPS, bolus priming system; BG, blood glucose.

Super-index refers to differences with no statistical significance in insulin and glycemia metrics between both strategies (P > .05).

In the nominal scenario, the UVA AID strategy led to 39 rebound hypoglycemia events. In contrast, when adding the HypoSafe module, all these events were avoided. In addition, results show no clinically significant differences in TIR (FCL = 80.4 ± 8.6 vs. FCL + HypoSafe = 79.8 ± 8.6), TBR (FCL = 2.0 ± 1.6 vs. FCL + HypoSafe = 1.7 ± 1.4), or TAR (FCL = 17.6 ± 8.2 vs. FCL + HypoSafe = 18.5 ± 8.3).

In the hypoglycemia-induced scenario, rebound hypoglycemia events were reduced from 55 to 7 with HypoSafe. Insulin delivery from basal and bolus doses was significantly reduced within the 30 minutes following a hypoglycemia event. Glucose outcome metrics show no clinically significant differences with and without HypoSafe.

Discussion

There are currently several strategies reported in literature and implemented in commercial devices that aim to alert or suspend insulin infusion for hypoglycemia prevention.^6,11,14,26 Typically, these strategies are based on glucose thresholds or short-term glucose predictions. However, once the subject treats a hypoglycemic event and BG levels rise rapidly, there can be an overreaction of insulin leading to rebound hypoglycemia or rollercoaster patterns.^22,27

In particular, a recent analysis on rebound hypoglycemia for subjects under multiple daily injections and continuous subcutaneous insulin infusion found 29.4% of hypoglycemic episodes were rebound events.²⁸ There was also an indication of behavioral overcompensation of carbohydrate intake when hypoglycemic and aggressiveness to correct resulting hyperglycemia.²⁸

This overreaction to rapid glucose increases has also been observed when using AID systems²² (see Supplemental Material). Features of the controller depending on glucose value or rate of change may trigger basal or bolus corrections after rescue carbohydrates, potentially leading to rebound hypoglycemia. This phenomenon is more common in FCL-enabled control strategies, which are designed to react more aggressively to positive glucose changes such as meals.

In this study, a HypoSafe module was developed to prevent rebound hypoglycemia caused by the controller overreacting to the effect of rescue carbohydrates. This safety layer is agnostic to the control strategy, ie, it acts as an outer layer that does not depend on the controller’s internal logic, it is simple to integrate, and only uses the last hour of CGM data as input, avoiding the need of requiring hypoglycemia treatment announcement and thus being suitable for FCL strategies.

The integration of this new control layer prevented rebound hypoglycemia events and lessened oscillatory glucose level behaviors in simulated experiments. The oscillatory response that can happen after a controller overcorrection following rescue carbohydrates is explicitly shown in Figure 3. This behavior is avoided by integrating HypoSafe as it restricts basal and bolus commands after the initial hypoglycemia treatment.

In addition, it has been observed that including HypoSafe has no clinically significant effect on glucose outcome metrics. This is advantageous since the module does not compensate for the reduction of hypoglycemia events by lowering TIR or increasing TAR. Hence, the performance achieved with the base controller design is maintained.

This work is limited to in-silico results comparing the performance of an AID system with and without HypoSafe. Conditions for hypoglycemia treatments were standardized for all virtual subjects (15 g every 20 minutes until BG >60 mg/dL), which can artificially reduce TBR for strategies with and without HypoSafe, but in free-living conditions, subject preferences and hypoglycemia awareness may differ from the testing scenario. This layer was integrated into the UVA AID system in a recent clinical trial (clinicaltrials.gov NCT05528770), where rebound hypoglycemia was not observed.

Conclusion

AID systems can increase the risk of rebound hypoglycemia after reacting to rescue carbohydrates. In this work, a HypoSafe module that can be easily integrated into an AID system has been developed to constrain insulin doses. HypoSafe is based on the minimum glucose measurement in the last hour and the current glucose concentration, avoiding the need of manual announcements. The proposed HypoSafe module was shown to be effective in reducing rebound hypoglycemia events without clinically affecting achieved TIR when combined with an advanced FCL system.

Supplemental Material

sj-docx-1-dst-10.1177_19322968231212205 – Supplemental material for Integration of a Safety Module to Prevent Rebound Hypoglycemia in Closed-Loop Artificial Pancreas Systems

sj-docx-1-dst-10.1177_19322968231212205.docx^{(204KB, docx)}

Supplemental material, sj-docx-1-dst-10.1177_19322968231212205 for Integration of a Safety Module to Prevent Rebound Hypoglycemia in Closed-Loop Artificial Pancreas Systems by María F. Villa-Tamayo, Patricio Colmegna and Marc D. Breton in Journal of Diabetes Science and Technology

Footnotes

Abbreviations: AID, automated insulin delivery; BG, blood glucose; BPS, bolus priming system; CGM, continuous glucose monitoring; FCL, full closed loop; FDA, Food and Drug Administration; HCL, hybrid closed loop; HMS, hyperglycemia mitigation system; MPC, model predictive control; T1D, type 1 diabetes; TDI, total daily insulin; TIR, time in range; TAR, time above range; TBR, time below range; UVA, University of Virginia.

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: M.F.V.-T., P.C., and M.D.B. receive research support and royalties from Dexcom handled by the University of Virginia’s Licensing and Ventures Group. M.D.B. reports consulting and speaker engagement from Dexcom, Tandem Diabetes Care, Sanofi, Roche, and Portal Insulin LLC.

Funding: The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by NIH grant 5R01DK129553.

ORCID iDs: María F. Villa-Tamayo Inline graphic https://orcid.org/0000-0002-0839-4070

Patricio Colmegna Inline graphic https://orcid.org/0000-0001-9074-8634

Marc D Breton Inline graphic https://orcid.org/0000-0001-7645-2693

Supplemental Material: Supplemental material for this article is available online.

References

1. Forlenza GP, Lal RA. Current status and emerging options for automated insulin delivery systems. Diabetes Technol Ther. 2022;24(5):362-371. doi: 10.1089/dia.2021.0514 [DOI] [PMC free article] [PubMed] [Google Scholar]
2. Nimri R, Phillip M, Kovatchev B. Decision support systems and closed-loop. Diabetes Technol Ther. 2022;24(s1):S58-S75. doi: 10.1089/dia.2022.2504 [DOI] [PubMed] [Google Scholar]
3. Moon SJ, Jung I, Park CY. Current advances of artificial pancreas systems: a comprehensive review of the clinical evidence. Diabetes Metab J. 2021;45(6):813-839. doi: 10.4093/dmj.2021.0177 [DOI] [PMC free article] [PubMed] [Google Scholar]
4. Garcia-Tirado J, Colmegna P, Villard O, et al. Assessment of meal anticipation for improving fully automated insulin delivery in adults with type 1 diabetes. Diabetes Care. 2023;46:1652-1658. doi: 10.2337/dc23-0119 [DOI] [PMC free article] [PubMed] [Google Scholar]
5. Blauw H, van Bon AC, Koops R, DeVries JH. Performance and safety of an integrated bihormonal artificial pancreas for fully automated glucose control at home. Diabetes Obes Metab. 2016;18(7):671-677. doi: 10.1111/dom.12663 [DOI] [PMC free article] [PubMed] [Google Scholar]
6. Urakami T. The advanced diabetes technologies for reduction of the frequency of hypoglycemia and minimizing the occurrence of severe hypoglycemia in children and adolescents with type 1 diabetes. J Clin Med. 2023;12(3). doi: 10.3390/jcm12030781 [DOI] [PMC free article] [PubMed] [Google Scholar]
7. Cryer PE. Hypoglycaemia: the limiting factor in the glycaemic management of type I and type II diabetes. Diabetologia. 2002;45(7):937-948. doi: 10.1007/s00125-002-0822-9 [DOI] [PubMed] [Google Scholar]
8. Bequette BW, Cameron F, Buckingham BA, Maahs DM, Lum J. Overnight hypoglycemia and hyperglycemia mitigation for individuals with type 1 diabetes: how risks can be reduced. IEEE Control Syst. 2018;38(1):125-134. doi: 10.1109/MCS.2017.2767119 [DOI] [Google Scholar]
9. Toffanin C, Aiello EM, Cobelli C, Magni L. Hypoglycemia prevention via personalized glucose-insulin models identified in free-living conditions. J Diabetes Sci Technol. 2019;13(6):1008-1016. doi: 10.1177/1932296819880864 [DOI] [PMC free article] [PubMed] [Google Scholar]
10. Hughes CS, Patek SD, Breton MD, Kovatchev BP. Hypoglycemia prevention via pump attenuation and red-yellow-green “traffic” lights using continuous glucose monitoring and insulin pump data. J Diabetes Sci Technol. 2010;4(5):1146-1155. doi: 10.1177/193229681000400513 [DOI] [PMC free article] [PubMed] [Google Scholar]
11. Puhr S, Derdzinski M, Welsh JB, Parker AS, Walker T, Price DA. Real-world hypoglycemia avoidance with a continuous glucose monitoring system’s predictive low glucose alert. Diabetes Technol Ther. 2019;21(4):155-158. doi: 10.1089/dia.2018.0359 [DOI] [PMC free article] [PubMed] [Google Scholar]
12. Turksoy K, Bayrak ES, Quinn L, Littlejohn E, Rollins D, Cinar A. Hypoglycemia early alarm systems based on multivariable models. Ind Eng Chem Res. 2013;52(35):12329-12336. doi: 10.1021/ie3034015 [DOI] [PMC free article] [PubMed] [Google Scholar]
13. Iacono F, Magni L, Toffanin C. Personalized LSTM-based alarm systems for hypoglycemia and hyperglycemia prevention. Biomed Signal Process Control. 2023;86. doi: 10.1016/j.bspc.2023.105167 [DOI] [Google Scholar]
14. Ly TT, Breton MD, Keith-Hynes P, et al. Overnight glucose control with an automated, unified safety system in children and adolescents with type 1 diabetes at diabetes camp. Diabetes Care. 2014;37(8):2310-2316. doi: 10.2337/dc14-0147 [DOI] [PMC free article] [PubMed] [Google Scholar]
15. Villa-Tamayo MF, Leon-Vargas F, Garcia-Jaramillo M, Rivadeneira PS. Glycemic control strategy based on an impulsive MPC with safety layer coupling for IOB limitation. IEEE Control Syst Lett. 2021;5(5):1669-1674. doi: 10.1109/LCSYS.2020.3042651 [DOI] [Google Scholar]
16. Fushimi E, Colmegna P, De Battista H, Garelli F, Sánchez-Peña R. Artificial pancreas: evaluating the ARG algorithm without meal announcement. J Diabetes Sci Technol. 2019;13(6):1035-1043. doi: 10.1177/1932296819864585 [DOI] [PMC free article] [PubMed] [Google Scholar]
17. Ellingsen C, Dassau E, Zisser H, et al. Safety constraints in an artificial pancreatic β cell: an implementation of model predictive control with insulin on board. J Diabetes Sci Technol. 2009;3(3):536-544. doi: 10.1177/193229680900300319 [DOI] [PMC free article] [PubMed] [Google Scholar]
18. Resalat N, El Youssef J, Reddy R, Castle J, Jacobs PG. Adaptive tuning of basal and bolus insulin to reduce postprandial hypoglycemia in a hybrid artificial pancreas. J Process Control. 2019;80:247-254. doi: 10.1016/j.jprocont.2019.05.018 [DOI] [Google Scholar]
19. Biester T, Kordonouri O, Holder M, et al. “Let the algorithm do the work”: reduction of hypoglycemia using sensor-augmented pump therapy with predictive insulin suspension (SmartGuard) in pediatric type 1 diabetes patients. Diabetes Technol Ther. 2017;19(3):173-182. doi: 10.1089/dia.2016.0349 [DOI] [PMC free article] [PubMed] [Google Scholar]
20. Yale JF, Paty B, Senior PA. Hypoglycemia. Can J Diabetes. 2018;42:S104-S108. doi: 10.1016/j.jcjd.2017.10.010 [DOI] [PubMed] [Google Scholar]
21. Pinsker JE, Bartee A, Katz M, et al. Predictive low-glucose suspend necessitates less carbohydrate supplementation to rescue hypoglycemia: need to revisit current hypoglycemia treatment guidelines. Diabetes Technol Ther. 2021;23:512-516. doi: 10.1089/dia.2020.0619 [DOI] [PMC free article] [PubMed] [Google Scholar]
22. Lee JB, Dassau E, Gondhalekar R, Seborg DE, Pinsker JE, Doyle FJ. Enhanced model predictive control (eMPC) strategy for automated glucose control. Ind Eng Chem Res. 2016;55(46):11857-11868. doi: 10.1021/acs.iecr.6b02718 [DOI] [PMC free article] [PubMed] [Google Scholar]
23. Garcia-Tirado J, Lv D, Corbett JP, Colmegna P, Breton MD. Advanced hybrid artificial pancreas system improves on unannounced meal response—in silico comparison to currently available system. Comput Methods Programs Biomed. 2021;211:106401. doi: 10.1016/j.cmpb.2021.106401 [DOI] [PubMed] [Google Scholar]
24. Visentin R, Campos-Náñez E, Schiavon M, et al. The UVA/Padova type 1 diabetes simulator goes from single meal to single day. J Diabetes Sci Technol. 2018;12(2):273-281. doi: 10.1177/1932296818757747 [DOI] [PMC free article] [PubMed] [Google Scholar]
25. American Diabetes Association. 6 glycemic targets: standards of medical care in diabetes—2019. Diabetes Care. 2019;42(suppl 1):S61-S70. doi: 10.2337/dc19-S006 [DOI] [PubMed] [Google Scholar]
26. Garg S, Brazg RL, Bailey TS, et al. Reduction in duration of hypoglycemia by automatic suspension of insulin delivery: the in-clinic aspire study. Diabetes Technol Ther. 2012;14(3):205-209. doi: 10.1089/dia.2011.0292 [DOI] [PubMed] [Google Scholar]
27. Acciaroli G, Welsh JB, Akturk HK. Mitigation of rebound hyperglycemia with real-time continuous glucose monitoring data and predictive alerts. J Diabetes Sci Technol. 2022;16(3):677-682. doi: 10.1177/1932296820982584 [DOI] [PMC free article] [PubMed] [Google Scholar]
28. Hansen KW, Bibby BM. Rebound hypoglycemia and hyperglycemia in type 1 diabetes [published online ahead of print May 3, 2023]. J Diabetes Sci Technol. 2023;3:19322968231168379. doi: 10.1177/19322968231168379 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

sj-docx-1-dst-10.1177_19322968231212205 – Supplemental material for Integration of a Safety Module to Prevent Rebound Hypoglycemia in Closed-Loop Artificial Pancreas Systems

sj-docx-1-dst-10.1177_19322968231212205.docx^{(204KB, docx)}

[bibr1-19322968231212205] 1. Forlenza GP, Lal RA. Current status and emerging options for automated insulin delivery systems. Diabetes Technol Ther. 2022;24(5):362-371. doi: 10.1089/dia.2021.0514 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr2-19322968231212205] 2. Nimri R, Phillip M, Kovatchev B. Decision support systems and closed-loop. Diabetes Technol Ther. 2022;24(s1):S58-S75. doi: 10.1089/dia.2022.2504 [DOI] [PubMed] [Google Scholar]

[bibr3-19322968231212205] 3. Moon SJ, Jung I, Park CY. Current advances of artificial pancreas systems: a comprehensive review of the clinical evidence. Diabetes Metab J. 2021;45(6):813-839. doi: 10.4093/dmj.2021.0177 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr4-19322968231212205] 4. Garcia-Tirado J, Colmegna P, Villard O, et al. Assessment of meal anticipation for improving fully automated insulin delivery in adults with type 1 diabetes. Diabetes Care. 2023;46:1652-1658. doi: 10.2337/dc23-0119 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr5-19322968231212205] 5. Blauw H, van Bon AC, Koops R, DeVries JH. Performance and safety of an integrated bihormonal artificial pancreas for fully automated glucose control at home. Diabetes Obes Metab. 2016;18(7):671-677. doi: 10.1111/dom.12663 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr6-19322968231212205] 6. Urakami T. The advanced diabetes technologies for reduction of the frequency of hypoglycemia and minimizing the occurrence of severe hypoglycemia in children and adolescents with type 1 diabetes. J Clin Med. 2023;12(3). doi: 10.3390/jcm12030781 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr7-19322968231212205] 7. Cryer PE. Hypoglycaemia: the limiting factor in the glycaemic management of type I and type II diabetes. Diabetologia. 2002;45(7):937-948. doi: 10.1007/s00125-002-0822-9 [DOI] [PubMed] [Google Scholar]

[bibr8-19322968231212205] 8. Bequette BW, Cameron F, Buckingham BA, Maahs DM, Lum J. Overnight hypoglycemia and hyperglycemia mitigation for individuals with type 1 diabetes: how risks can be reduced. IEEE Control Syst. 2018;38(1):125-134. doi: 10.1109/MCS.2017.2767119 [DOI] [Google Scholar]

[bibr9-19322968231212205] 9. Toffanin C, Aiello EM, Cobelli C, Magni L. Hypoglycemia prevention via personalized glucose-insulin models identified in free-living conditions. J Diabetes Sci Technol. 2019;13(6):1008-1016. doi: 10.1177/1932296819880864 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr10-19322968231212205] 10. Hughes CS, Patek SD, Breton MD, Kovatchev BP. Hypoglycemia prevention via pump attenuation and red-yellow-green “traffic” lights using continuous glucose monitoring and insulin pump data. J Diabetes Sci Technol. 2010;4(5):1146-1155. doi: 10.1177/193229681000400513 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr11-19322968231212205] 11. Puhr S, Derdzinski M, Welsh JB, Parker AS, Walker T, Price DA. Real-world hypoglycemia avoidance with a continuous glucose monitoring system’s predictive low glucose alert. Diabetes Technol Ther. 2019;21(4):155-158. doi: 10.1089/dia.2018.0359 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr12-19322968231212205] 12. Turksoy K, Bayrak ES, Quinn L, Littlejohn E, Rollins D, Cinar A. Hypoglycemia early alarm systems based on multivariable models. Ind Eng Chem Res. 2013;52(35):12329-12336. doi: 10.1021/ie3034015 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr13-19322968231212205] 13. Iacono F, Magni L, Toffanin C. Personalized LSTM-based alarm systems for hypoglycemia and hyperglycemia prevention. Biomed Signal Process Control. 2023;86. doi: 10.1016/j.bspc.2023.105167 [DOI] [Google Scholar]

[bibr14-19322968231212205] 14. Ly TT, Breton MD, Keith-Hynes P, et al. Overnight glucose control with an automated, unified safety system in children and adolescents with type 1 diabetes at diabetes camp. Diabetes Care. 2014;37(8):2310-2316. doi: 10.2337/dc14-0147 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr15-19322968231212205] 15. Villa-Tamayo MF, Leon-Vargas F, Garcia-Jaramillo M, Rivadeneira PS. Glycemic control strategy based on an impulsive MPC with safety layer coupling for IOB limitation. IEEE Control Syst Lett. 2021;5(5):1669-1674. doi: 10.1109/LCSYS.2020.3042651 [DOI] [Google Scholar]

[bibr16-19322968231212205] 16. Fushimi E, Colmegna P, De Battista H, Garelli F, Sánchez-Peña R. Artificial pancreas: evaluating the ARG algorithm without meal announcement. J Diabetes Sci Technol. 2019;13(6):1035-1043. doi: 10.1177/1932296819864585 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr17-19322968231212205] 17. Ellingsen C, Dassau E, Zisser H, et al. Safety constraints in an artificial pancreatic β cell: an implementation of model predictive control with insulin on board. J Diabetes Sci Technol. 2009;3(3):536-544. doi: 10.1177/193229680900300319 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr18-19322968231212205] 18. Resalat N, El Youssef J, Reddy R, Castle J, Jacobs PG. Adaptive tuning of basal and bolus insulin to reduce postprandial hypoglycemia in a hybrid artificial pancreas. J Process Control. 2019;80:247-254. doi: 10.1016/j.jprocont.2019.05.018 [DOI] [Google Scholar]

[bibr19-19322968231212205] 19. Biester T, Kordonouri O, Holder M, et al. “Let the algorithm do the work”: reduction of hypoglycemia using sensor-augmented pump therapy with predictive insulin suspension (SmartGuard) in pediatric type 1 diabetes patients. Diabetes Technol Ther. 2017;19(3):173-182. doi: 10.1089/dia.2016.0349 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr20-19322968231212205] 20. Yale JF, Paty B, Senior PA. Hypoglycemia. Can J Diabetes. 2018;42:S104-S108. doi: 10.1016/j.jcjd.2017.10.010 [DOI] [PubMed] [Google Scholar]

[bibr21-19322968231212205] 21. Pinsker JE, Bartee A, Katz M, et al. Predictive low-glucose suspend necessitates less carbohydrate supplementation to rescue hypoglycemia: need to revisit current hypoglycemia treatment guidelines. Diabetes Technol Ther. 2021;23:512-516. doi: 10.1089/dia.2020.0619 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr22-19322968231212205] 22. Lee JB, Dassau E, Gondhalekar R, Seborg DE, Pinsker JE, Doyle FJ. Enhanced model predictive control (eMPC) strategy for automated glucose control. Ind Eng Chem Res. 2016;55(46):11857-11868. doi: 10.1021/acs.iecr.6b02718 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr23-19322968231212205] 23. Garcia-Tirado J, Lv D, Corbett JP, Colmegna P, Breton MD. Advanced hybrid artificial pancreas system improves on unannounced meal response—in silico comparison to currently available system. Comput Methods Programs Biomed. 2021;211:106401. doi: 10.1016/j.cmpb.2021.106401 [DOI] [PubMed] [Google Scholar]

[bibr24-19322968231212205] 24. Visentin R, Campos-Náñez E, Schiavon M, et al. The UVA/Padova type 1 diabetes simulator goes from single meal to single day. J Diabetes Sci Technol. 2018;12(2):273-281. doi: 10.1177/1932296818757747 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr25-19322968231212205] 25. American Diabetes Association. 6 glycemic targets: standards of medical care in diabetes—2019. Diabetes Care. 2019;42(suppl 1):S61-S70. doi: 10.2337/dc19-S006 [DOI] [PubMed] [Google Scholar]

[bibr26-19322968231212205] 26. Garg S, Brazg RL, Bailey TS, et al. Reduction in duration of hypoglycemia by automatic suspension of insulin delivery: the in-clinic aspire study. Diabetes Technol Ther. 2012;14(3):205-209. doi: 10.1089/dia.2011.0292 [DOI] [PubMed] [Google Scholar]

[bibr27-19322968231212205] 27. Acciaroli G, Welsh JB, Akturk HK. Mitigation of rebound hyperglycemia with real-time continuous glucose monitoring data and predictive alerts. J Diabetes Sci Technol. 2022;16(3):677-682. doi: 10.1177/1932296820982584 [DOI] [PMC free article] [PubMed] [Google Scholar]

[bibr28-19322968231212205] 28. Hansen KW, Bibby BM. Rebound hypoglycemia and hyperglycemia in type 1 diabetes [published online ahead of print May 3, 2023]. J Diabetes Sci Technol. 2023;3:19322968231168379. doi: 10.1177/19322968231168379 [DOI] [PMC free article] [PubMed] [Google Scholar]

PERMALINK

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

María F Villa-Tamayo, MS

Patricio Colmegna, PhD

Marc D Breton, PhD