1. What a closed loop is, in three points

A closed loop is a system that runs the cycle “measure → decide → deliver → measure again” without a person stepping in each time (our commentary). In diabetes it is called automated insulin delivery (AID) and sometimes an artificial pancreas.

• What it is made of: FDA regulations divide it into an iCGM that measures glucose (21 CFR 862.1355), an iAGC (interoperable automated glycemic controller, 21 CFR 862.1356) that calculates doses and issues commands, and an ACE pump (alternate controller enabled infusion pump, 21 CFR 880.5730) that receives commands and infusesSourced
• Is the loop fully closed: most current products are “hybrid” designs that leave user actions in place, for example at mealtimes. The FDA document for Control-IQ says mealtime boluses are delivered manually by the userSourced
• Where materials come in: the infusion pathway (reservoir, tubing, cannula), the adhesive that holds devices on the skin, sensors that last days to two weeks, and the stability of insulin, which is a drug. Any of these can stop the loop even if the algorithm is right (our commentary)

2. Three components and how they connect: sensor, algorithm, pump

The regulation text sets out agreements between the components in detailSourced.

• iAGC (862.1356): a device that automatically calculates drug doses from inputs such as glucose and commands a connected pump to deliver them. As a rule, glucose input comes from an iCGM (862.1355) and commands go to an ACE pump (880.5730). Documentation is required showing that safe therapy is maintained when communication is interrupted, lost or re-established [Reference 2].
• ACE pump (880.5730): delivers basal and bolus infusion, and communicates securely with external automated dosing systems to receive, execute and confirm commands [Reference 3].
• iCGM (862.1355): a Class II CGM designed to connect with other devices; its special controls include meeting agreement-rate criteria by range in an adult clinical study [Reference 1]. See our explainer on continuous glucose monitoring for details.

3. What the control algorithm does (as FDA documents describe it)

The inside of an algorithm is confidential, but FDA review documents give an outline of how it behaves. We introduce only the description of operation here, and do not touch settings or usage.

All Sourced (FDA documents [References 5, 6, 8, 9]). Indicated ages and compatible devices differ by product and version and are set in the labelling.

Every one of these systems depends on the accuracy and lag of the sensor value. From the control side, sensor error and lag flow directly into dosing decisions. That is why the iCGM special controls require not only agreement rates by range but also freedom from rate-of-change direction errors and data gaps [Reference 1] (our commentary). Doyle and colleagues' review sets out the design objectives, variables and challenges involved in developing an artificial pancreas [Reference 19].

4. FDA classification: from an integrated PMA to component-by-component classes

At first, the FDA recognised automated insulin delivery by granting one PMA (approval) for the whole system. It then moved to classes for each component, which are combined. FDA “approval”, “grant” and “clearance” are separate procedures.

All Sourced (21 CFR 862.1356 and 880.5730 [References 2, 3] and the FDA product classification database [Reference 17]). For children, the iAGC labelling must carry a contraindication unless clinical performance data or other information show it can be used safely.

The first De Novo (DEN180058, Tandem t:slim X2) recognised the pump on its own as “a component that can receive commands from external automated dosing software”. According to the review document, this pump shares its hardware and much of its software with already approved pumps, and only some software was modifiedSourced [Reference 4]. The next, DEN190034 (Control-IQ), recognised a software-only iAGC and named the Dexcom G6 as the compatible iCGM at the time of evaluationSourced [Reference 5]. This “class designed for combination” has led to the 510(k)s that companies have filed since (our commentary).

5. A materials engineer's view (1): wetted materials, sterilisation and insulin stability

View the original SVG schematic, source notes and numerical details

6. Our calculation: how small one delivery step is, and the volume before an occlusion is detected

7. Cannula and infusion-set failures: published numbers

The last exit of the loop is the thin cannula under the skin. If it blocks, shifts or kinks, insulin does not arrive even though the pump and the algorithm are working normally.

• An example from a review document: in the six-month randomised Control-IQ comparison (168 people) there was one diabetic ketoacidosis, whose cause is recorded as an infusion set failure. Hyperglycaemia or ketosis events meeting the reporting criteria numbered 13 in the closed-loop arm (112 people) and 2 in the control (sensor-augmented pump) arm (56 people)Sourced [Reference 5].
• Types of failure: an expert panel report lists the sources of infusion-set failure as kinking, dislocation or leakage of the cannula, reservoir leakage, a loose connection between reservoir and cannula, insulin precipitation in the cannula, and occlusions that may or may not trigger pump alarmsSourced [Reference 12].
• How little is known: a review by Heinemann and Krinelke says published knowledge about infusion sets is relatively small and that no prospective clinical study on current infusion sets and insulin formulations gives representative data on the type and frequency of their problemsSourced [Reference 13].

A study that quantified frequency is that of Kanapka and colleagues. It analysed 22,741 infusion sets used by 263 participants aged 6 to 72 in two clinical trialsSourced [Reference 10].

• Sets removed before 72 hours in association with prolonged hyperglycaemia numbered 748 (3.3%). Under a less restrictive definition (above 300 mg/dL at the time of removal) the figure was 1,688 (7.4%)Sourced [Reference 10].
• By age, 1.9% for those 18 and over, 5.8% for 14 to 17 and 4.4% for 6 to 13. By set type, 4.0% for 90° Teflon sets, 1.9% for steel sets and 1.3% for angled Teflon setsSourced [Reference 10].
• The authors say such changes occur about four times a year on averageSourced [Reference 10]. Changing every 3 days uses about 122 sets a year (365 / 3); multiplied by 0.033 that is about fourOur calculation.

This figure rests on an operational definition, “removed after the sensor showed prolonged high glucose”, and does not separate causes such as kinking, occlusion or dislodgement. Seen through a materials lens, the result suggests that cannula material, shape and insertion angle may affect failure frequency, but identifying causes would take device-by-device studies (our commentary).

8. A materials engineer's view (2): three consumables on different cycles, and the burden of adhesion

All Sourced (FDA documents [References 4 to 7, 18, 20 to 22]). Updated with each product and version.

9. When communication drops: what design must provide

The loop runs only when three devices are connected wirelessly. The regulation text sets out what must happen when the connection is lostSourced [References 2, 3].

• ACE pump: it must include measures to keep therapy safe when communication with a digitally connected alternate controller is interrupted, lost or re-established (for example, reverting to a pre-programmed safe delivery rate), and validation must show that events during the loss (commands, malfunctions, occlusions and so on) are handled and logged appropriately.
• iAGC: documentation is needed showing that safe therapy is maintained across interruption, loss and re-establishment, and critical events such as commands and confirmations, malfunctions, alarms and their acknowledgements, and connectivity events must be stored and accessible for an adequate period.
• Both: the interface requirements include secure authentication (pairing) with the other device, sharing of state information (battery, reservoir level, sensor use life, errors) and continuing to operate safely when data outside the specified bounds is received.

From a materials viewpoint, hardware reliability (batteries, antennas, waterproofing of housings, connection to the sensor) bears directly on how easily communication drops (our commentary). The special controls for ACE pumps also require reliability of pump function over the use life, and testing of electrical safety, electromagnetic compatibility and the safety and availability of the radio linkSourced [Reference 3].

10. Open problems, and what this article could not confirm

(1) The specific materials of each company's infusion sets, cannulas and adhesives

Among the FDA documents checked for this article, only the Omnipod 5 510(k) listed fluid-pathway materials. The specific materials of cannulas and adhesives for the t:slim X2, MiniMed, iLet and other products could not be confirmed, and the Omnipod 5 document does not say which material is used in which part.

(2) Algorithm internals, and comparisons between systems

The inside of the algorithms (models, parameters, the handling of targets) could not be confirmed in published documents, so only outlines of operation are given. Comparisons of clinical results between systems are not covered, because populations and conditions differ.

(3) Interaction of insulin formulation additives with materials

Interactions between additives such as preservatives (m-cresol, phenol) or zinc and the pathway materials were not shown in detail in the documents checked. The swine study [Reference 11] gives the result under one set of experimental conditions.

(4) Regulation in other regions, and systems built by individuals

Classification in Europe (MDR) and Japan (PMDA), and automated delivery systems that individuals build, for example with open-source tools, were neither checked nor discussed.

11. Glossary

12. References (primary sources)

1. U.S. Food and Drug Administration (FDA; eCFR) “21 CFR 862.1355 Integrated continuous glucose monitoring system” — www.ecfr.gov
2. FDA (eCFR) “21 CFR 862.1356 Interoperable automated glycemic controller” — www.ecfr.gov
3. FDA (eCFR) “21 CFR 880.5730 Alternate controller enabled infusion pump” — www.ecfr.gov
4. FDA “DEN180058 Evaluation of Automatic Class III Designation for t:slim X2 insulin pump with interoperable technology: Decision Summary” — www.accessdata.fda.gov
5. FDA “DEN190034 Evaluation of Automatic Class III Designation for Control-IQ Technology: Decision Memorandum” — www.accessdata.fda.gov
6. FDA “PMA P160017 FDA Summary of Safety and Effectiveness Data (MiniMed 670G System)”, approved 28 September 2016 — www.accessdata.fda.gov
7. FDA “K203768 510(k) summary and clearance letter (Omnipod 5 ACE Pump)”, 27 January 2022 — www.accessdata.fda.gov
8. FDA “K203774 510(k) summary and clearance letter (SmartAdjust technology)”, 27 January 2022 — www.accessdata.fda.gov
9. FDA “K220916 510(k) summary and clearance letter (iLet Dosing Decision Software)”, 19 May 2023 — www.accessdata.fda.gov
10. Kanapka L et al. “Insulin Pump Infusion Set Failures Associated with Prolonged Hyperglycemia: Frequency and Relationship to Age and Type of Infusion Set During 22,741 Infusion Set Wears”, Diabetes Technol Ther 24:396 (2022) — doi.org
11. Swinney MR et al. “Insulin, Not the Preservative m-cresol, Instigates Loss of Infusion Site Patency Over Extended Durations of CSII in Diabetic Swine”, J Pharm Sci 110:1418 (2021) — doi.org
12. Deiss D et al. “Insulin Infusion Set Use: European Perspectives and Recommendations”, Diabetes Technol Ther 18:517 (2016) (PMC) — pmc.ncbi.nlm.nih.gov
13. Heinemann L, Krinelke L. “Insulin Infusion Set: The Achilles Heel of Continuous Subcutaneous Insulin Infusion”, J Diabetes Sci Technol 6:954 (2012) — doi.org
14. FDA “K223846 510(k) summary and clearance letter (iLet ACE Pump)”, 19 May 2023 — www.accessdata.fda.gov
15. FDA “K251032 510(k) summary and clearance letter (MiniMed 780G insulin pump)”, 1 July 2025 — www.accessdata.fda.gov
16. FDA “K251217 510(k) summary and clearance letter (SmartGuard technology; Predictive Low Glucose technology)”, 29 August 2025 — www.accessdata.fda.gov
17. FDA Product classification database, product code OZP “Automated Insulin Dosing Device System, Single Hormonal Control” (via the openFDA API) — api.fda.gov
18. FDA “K233537 510(k) summary and clearance letter (FreeStyle Libre 2 / Libre 3)”, 23 April 2024 — www.accessdata.fda.gov
19. Doyle FJ 3rd et al. “Closed-loop artificial pancreas systems: engineering the algorithms”, Diabetes Care 37:1191 (2014) — doi.org
20. FDA “DEN170088 Dexcom G6 Decision Summary”, 27 March 2018 — www.accessdata.fda.gov
21. FDA “K234133 510(k) summary and clearance letter (Dexcom G7)”, 26 February 2024 — www.accessdata.fda.gov
22. FDA “K243214 510(k) summary and clearance letter (Dexcom G7 15 Day)”, 9 April 2025 — www.accessdata.fda.gov

13. Claim-to-source audit

Last updated 3 October 2026. Sources are limited to primary material (FDA regulation text, review documents, 510(k) summaries and the product classification database; peer-reviewed papers). No market-size or market-share estimates are used. Because the article includes structural readings and materials-design interpretations, those are marked as Commentary and kept separate from sourced fact. The specific materials of infusion sets, cannulas and adhesives (other than for Omnipod 5), algorithm internals, clinical comparisons between systems, interactions of insulin additives with materials, and regulation in other regions are not covered, because they could not be confirmed in published primary sources. Figs. 1, 2 and 3 are concept diagrams, Fig. 4 includes our calculation, and the hero image and Fig. 5 are AI-generated images; none of them shows a real product, cross-section or measured data. This article explains technology. It gives no advice on diagnosis, treatment, insulin doses or settings, and it does not recommend any product.