Is there a patch that senses glucose and doses insulin?

Automated insulin delivery exists and works, but it is built from separate parts. In the US the FDA regulates the parts as three device types: an integrated continuous glucose monitor (iCGM, 21 CFR 862.1355), an alternate controller enabled insulin pump (ACE pump, 21 CFR 880.5730) and an interoperable automated glycemic controller, the dosing algorithm (21 CFR 862.1356). The closest published attempt at one device that senses and doses is an investigational patch pump tested in 18 adults, and in that study people still used a separate CGM to make treatment decisions. Patches that release insulin in response to glucose without electronics have worked in mice and minipigs, not yet in published human trials.

How does closed-loop insulin work today?

A sensor reads glucose in the fluid under the skin, an algorithm decides the dose, and a pump delivers it. The FDA's own description of the iLet Bionic Pancreas shows the split. In May 2023 it cleared two devices, the iLet ACE Pump and the iLet Dosing Decision Software, which "along with a compatible FDA-cleared integrated continuous glucose monitor (iCGM), will form a new system". The software "determines and commands" basal and correction doses from the iCGM's readings (FDA, 19 May 2023).

The approach has randomised evidence behind it. In the six-month iDCL trial, 168 people with type 1 diabetes were randomised 2:1 to a closed-loop system or a sensor-augmented pump. Time in the 70–180 mg/dL range rose from 61% to 71% with closed loop and stayed at 59% in the control group, an adjusted difference of 11 percentage points (Brown et al., N Engl J Med 2019).

What would a single sense-and-dose patch have to prove?

All three sets of FDA special controls, at once, in one body. Each regulation is Class II with its own requirements. A few that a one-patch design cannot dodge:

• Sensor accuracy (iCGM). Below 70 mg/dL, the share of readings within ±15 mg/dL of a lab reference must have a lower one-sided 95% confidence bound above 85%. From 70 to 180 mg/dL, readings within ±15% must clear 70%; above 180 mg/dL, 80% (21 CFR 862.1355). Performance must hold "throughout the sensor wear period".
• Delivery accuracy and hazard detection (ACE pump). Delivery accuracy must hold under foreseeable changes in temperature and pressure; the pump must detect hazards such as occlusions within a clinically relevant time; and the drug's stability must be tested under conditions such as temperature and light (21 CFR 880.5730).
• Clinical performance and safe failure (controller). The algorithm needs clinical data in the intended population, a human-factors study, and validated behaviour that keeps therapy safe "when communication with digitally connected devices is interrupted, lost, or re-established" (21 CFR 862.1356).

A worked example of why one patch is harder, not easier. In today's systems the sensor and the pump are separate devices, each tested against its own rules. Put both in one patch and the iCGM accuracy bounds above have to be met at the same site where insulin is delivered, for the whole wear period, while the pump's rules on delivery accuracy, occlusion and drug stability still apply to the same object. That is Micropatch's reading of the rules, not an FDA statement; the research brief sets out the trade-offs (research brief, section 08).

What is the closest thing that exists?

An investigational all-in-one patch pump. PharmaSens reported a first-in-human feasibility study of its niia patch pump with a Pacific Diabetes Technologies glucose sensor built into the insulin cannula. It was an early, single-arm study in 18 adults with type 1 diabetes; 15 wore the device for 72 hours. Participants used a separate Dexcom G6 CGM and a blood glucose meter for treatment decisions and took manual bolus doses, so it was not run as a closed loop. The primary endpoint was the device's glucose accuracy against a lab analyser. This is a company release about an investigational device (PharmaSens release, 10 April 2026).

Glucose-responsive "smart insulin patches", in animals. In 2015 a microneedle patch loaded with insulin and glucose oxidase in glucose-responsive vesicles regulated blood glucose in a mouse model of type 1 diabetes (Yu et al., PNAS 2015). In 2020 a single removable microneedle patch with a glucose-responsive polymer regulated blood glucose in diabetic mice and minipigs; in minipigs over 25 kg, control lasted more than 20 hours with patches of about 5 cm² (Yu et al., Nat Biomed Eng 2020). Several authors of the 2020 paper disclosed patents and roles at a company developing the approach. Neither study was in people.

Where does Micropatch fit?

Micropatch is a concept for a patch you would change once a day that reads glucose and doses insulin. It is a way of asking what such a device would need, with every figure traced to a source in the research brief. It is not a product, has no trial behind it, and is not a substitute for any treatment.

FAQ