The Artificial Pancreas (Closed-Loop Systems)
An artificial pancreas, or closed-loop system, pairs a glucose sensor and an insulin pump with control software that adjusts insulin delivery [1]. For most of its history, it has leaned on the fingerstick. The first closed loops, in the 1970s, skipped the meter only because they measured glucose in blood drawn from a vein at a hospital bedside [1]. The first hybrid closed loop the FDA approved, the MiniMed 670G in 2016 [2], came with its own meter and needed a fingerstick calibration at least every 12 hours [3]. Below we follow the meter through that history, up to the factory-calibrated sensors of October 2026, and look at what published studies count. Our CGM hub covers these systems from the sensor side in Closed-Loop Systems (Artificial Pancreas).
Bedside machines that read blood directly
The earliest systems measured glucose intravenously and infused insulin and glucose the same way [1]. Kadish reported pioneering work in 1964 [1]. Then five research teams published closed-loop results between 1974 and 1978, and one of those designs became the first commercial device, the Biostator, in 1977 [1]. The intravenous route kept it in the hospital, but it proved that a machine could hold glucose in check [1]. No test strip was needed, because the device sampled blood itself.
1999 to 2016: sensors that needed the meter
The move toward a wearable loop began with a sensor under the skin. In June 1999 the FDA approved MiniMed's Continuous Glucose Monitoring System, which recorded glucose in interstitial fluid (the fluid around cells) so a clinician could download and review it later [4]. Its approval order said the data were meant to supplement, not replace, readings from a home glucose meter [4]. That sensor set off a wave of work on systems that both sensed glucose and delivered insulin under the skin, and in September 2006 JDRF launched its Artificial Pancreas Project to fund closed-loop research [1].
Calibration kept the meter in the loop. Interstitial glucose can differ markedly from blood glucose after a meal or during a low, so as of 2011 CGM devices still needed calibrating with one or more blood samples a day [1]. Even the Dexcom G5 Mobile, which FDA approved in December 2016 to replace fingerstick testing for treatment decisions, took blood glucose calibrations entered on its receiver or app [5]. The G5 came before the G6 [6] and the G7 sensors Dexcom offers as of October 2026 [7].
2016: the first hybrid closed loop shipped with a meter
FDA approved the MiniMed 670G on September 28, 2016, and calls it the first hybrid closed-loop system [2]. "Hybrid" meant the pump adjusted background (basal) insulin on its own, while meal boluses stayed the user's job [3]. Here's what surprised us in FDA's approval summary: the meter was a listed part of the system. The 670G included a Contour Next Link 2.4 meter, used with Contour Next test strips, that sent readings to the pump by radio [3].
The sensor leaned hard on those strips. FDA's summary describes calibrating the Guardian Sensor 3 at least every 12 hours, or three to four times a day [3]. Its sensor values weren't meant for therapy adjustments; those were to rest on home meter readings, and in Auto Mode, meter readings were used for any correction bolus [3]. The labeling also said pump therapy wasn't recommended for people unwilling or unable to do at least four blood glucose tests a day [3]. The 670G has since been superseded by the MiniMed 780G, which began a broad US launch with Abbott's Instinct sensor in December 2025 [8].
2018 and 2019: calibration becomes optional
The Dexcom G6 made the calibration fingerstick optional. FDA authorized it through a De Novo request, the pathway for a new kind of device, on March 27, 2018 [6]. Each sensor came with a code. Once the user entered it, no calibration was needed for the full 10-day wear; skip the code, and the G6 asked for two fingersticks at startup and one every 24 hours after that [6]. FDA classed the G6 as an integrated continuous glucose monitoring system (iCGM), cleared to send readings to automated insulin dosing systems [6]. Dexcom said it would no longer manufacture the G6 after July 1, 2026, and it points G6 users to the G7 or G7 15 Day [7].
The loop closed around that sensor in December 2019, when FDA authorized Tandem's Control-IQ technology with the G6 as its compatible iCGM [9]. FDA's decision memo also notes that iCGM point values are "typically less accurate" than meter results [9]. Tandem's newer Control-IQ+ was cleared in February 2025 with Control-IQ as its predicate device, and its 510(k) summary states that the algorithm design is unchanged [10].
Two sensor lines that work with automated insulin delivery (AID) systems as of October 2026 skip calibration by default. The Dexcom G7 is factory calibrated, with optional fingerstick calibration [11]. The FreeStyle Libre 3 Plus and 2 Plus sensors are factory calibrated, need no fingerstick calibration, and are cleared for use with AID systems [12]. What each system does when its sensor stops sending readings is covered in our article on Closed-Loop Control Algorithms (Artificial Pancreas).
What studies report about fingersticks on AID
The calibration years left a record of what fingersticks cost users. In a Stanford clinic cohort of 79 people starting the 670G, 26 (33%) had stopped using Auto Mode by 12 months [13]. Sensor issues were the most common reason, and the authors named alarms and sensor calibration as a major patient concern [13]. A Colorado study followed 92 young people starting the same system. Within six months, 28 (30%) had stopped, and those who quit ranked calibrations among the hardest parts, alongside alarms and the time the system demanded [14].
On the 670G, strips and automation went together. A year-long study of 276 users found that four or more blood glucose checks a day was one of the factors linked with reaching the study's target for time spent in closed-loop mode [15].
Optional calibration changed the count. An analysis of pivotal-trial data, written by a Barbara Davis Center researcher and three Medtronic employees, compared the 780G with the Guardian 4 sensor against the 670G with Guardian Sensor 3. Fingersticks fell by 6.2 a day in adults and 6.9 a day in children [16]. A smaller single-center study in Japan followed 21 adults who moved from the 770G with Guardian Sensor 3 to the 780G with Guardian 4. Their calibrations, each one a fingerstick typed into the pump, dropped from 3.6 to 0.8 a day over three months [17].
Every count in this section comes from Medtronic systems. We didn't find a comparable published count for AID users on factory-calibrated Dexcom or Abbott sensors, so how often they still reach for a meter is an open question. How often you should check with a meter on any of these systems is a decision to make with your care team.
References
- Artificial Pancreas: Past, Present, Future (Cobelli C, Renard E, Kovatchev B. Diabetes. 2011;60(11):2672-2682)Source
- The Artificial Pancreas Device System (U.S. Food and Drug Administration)Source
- Summary of Safety and Effectiveness Data: MiniMed 670G System, PMA P160017 (U.S. Food and Drug Administration)Source
- PMA P980022: MiniMed Continuous Glucose Monitoring System, original approval and approval order statement (U.S. Food and Drug Administration, openFDA PMA record)Source
- PMA P120005/S041: Dexcom G5 Mobile Continuous Glucose Monitoring System, approval order statement (U.S. Food and Drug Administration, openFDA PMA record)Source
- De Novo DEN170088 Decision Summary: Dexcom G6 Continuous Glucose Monitoring System (U.S. Food and Drug Administration)Source
- Dexcom G6 transition: upgrade to Dexcom G7 or G7 15 Day (Dexcom)Source
- Medtronic begins U.S. commercial launch of the MiniMed 780G system with the Instinct sensor, made by Abbott (MiniMed press release)Source
- De Novo DEN190034 Decision Memorandum: Control-IQ Technology, interoperable automated glycemic controller (U.S. Food and Drug Administration)Source
- 510(k) K243823: Control-IQ+ technology, interoperable automated glycemic controller (Tandem Diabetes Care; U.S. Food and Drug Administration)Source
- 510(k) Decision Summary K213919: Dexcom G7 Continuous Glucose Monitoring System (U.S. Food and Drug Administration)Source
- 510(k) Decision Summary K233537: FreeStyle Libre 3 and FreeStyle Libre 2 systems (FreeStyle Libre 3 Plus and 2 Plus sensors) (U.S. Food and Drug Administration)Source
- One Year Clinical Experience of the First Commercial Hybrid Closed-Loop System (Lal RA, Basina M, Maahs DM, et al. Diabetes Care. 2019;42(12):2190-2196)Source
- Real world hybrid closed-loop discontinuation: Predictors and perceptions of youth discontinuing the 670G system in the first 6 months (Messer LH, Berget C, Vigers T, et al. Pediatr Diabetes. 2020;21(2):319-327)Source
- Real-world performance of hybrid closed loop in youth, young adults, adults and older adults with type 1 diabetes: Identifying a clinical target for hybrid closed-loop use (Berget C, Akturk HK, Messer LH, et al. Diabetes Obes Metab. 2021;23(9):2048-2057)Source
- Reducing Diabetes Burden in Medtronic's Automated Insulin Delivery Systems (Forlenza GP, Dai Z, Niu F, Shin JJ. Diabetes Technol Ther. 2024;26(S3):7-16)Source
- Efficacy of MiniMed 780G on glucose management, sleep quality, and psychological aspects in persons with type 1 diabetes treated with MiniMed 770G: A 3-month prospective single-center observational study (Akiyama T, Orime K, Akamatsu R, et al. J Diabetes Investig. 2026;17(1):165-173)Source