CGM Sensor Technology Mechanisms

Every CGM sensor does the same basic job. It sits in the fluid just under the skin, turns the glucose it finds there into a signal, and hands that signal to software that converts it into a number every few minutes [1]. What differs is how the signal is made. As of October 2026, Dexcom's G7 sensors and Abbott's FreeStyle Libre Plus sensors use an enzyme called glucose oxidase on an electrode [1][2], while the Eversense 365, an implant from Senseonics, reads glucose with a fluorescent chemical and light [3].

Where the sensor sits and what it actually measures

A CGM doesn't measure blood. Its sensor sits in interstitial fluid, the watery layer that surrounds cells beneath the skin. The Dexcom G7 15 Day, for example, reports a new value from that fluid every 5 minutes across a range of 40 to 400 mg/dL [1].

Glucose has to travel out of the capillaries before the sensor can see it. When Mayo Clinic researchers traced labeled glucose in eight healthy adults after an overnight fast, it appeared in the interstitial fluid 5 to 6 minutes after it appeared in the blood [4]. They concluded that this delay shouldn't limit sensor accuracy overnight or in a fasting state [4]. Labeling is more cautious: FDA's decision summary for the Dexcom G6 says the device shouldn't be used for treatment decisions while glucose is rising or falling rapidly [5]. How to act on any reading is a decision for your care team.

The enzyme electrode: how Dexcom sensors work

FDA's review of the Dexcom G7 15 Day describes the chemistry in a short chain [1]. Glucose oxidase on the sensor probe oxidizes glucose. That reaction produces hydrogen peroxide, and the hydrogen peroxide generates an electrical current in proportion to the glucose concentration. The transmitter then runs that current through an algorithm to produce a value in mg/dL.

FDA described the same principle for the Dexcom G6 in 2018 [5]. The G6 matters for history because its De Novo authorization created the federal category for integrated CGMs, or iCGMs: sensors built to send readings to other devices, including automated insulin dosing systems [5][6]. Dexcom has since replaced the G6 with its G7 sensors and says it will no longer manufacture the G6 after July 1, 2026 [7].

Abbott's variation: a wired enzyme

Abbott's FreeStyle Libre sensors also start with glucose oxidase, but they move electrons a different way. In FDA's review of the Libre 3 Plus and Libre 2 Plus sensors, the enzyme oxidizes glucose and passes electrons to an electrode, producing a current proportional to glucose [2]. A 2017 paper by Abbott scientists fills in the rest. The enzyme is crosslinked into a polymer with an osmium compound that acts as a mediator, a molecule that shuttles electrons from the enzyme to the electrode [8]. Abbott calls this Wired Enzyme technology, and the authors say its low operating voltage means fewer other substances react at the electrode [8]. They work for Abbott, so read that as the manufacturer explaining its own design.

As of October 2026, the FreeStyle Libre 3 system uses the Libre 3 Plus sensor, worn on the back of the upper arm for up to 15 days after a 60-minute warm-up [9].

Reading glucose with light: Eversense 365

The Eversense 365 skips the enzyme entirely. A healthcare provider inserts the sensor under the skin, and its surface carries a fluorescent indicator molecule [3]. Glucose binds to that indicator reversibly, the binding makes it fluoresce more strongly, and an optical system in the sensor measures the change. The implant is powered by radio frequency from a transmitter worn outside the body, and that transmitter calculates the glucose value and sends it to a phone app [3]. FDA cleared the system in September 2024 for up to one year of wear in adults 18 and older [3]. Our page on implantable fluorescence sensors covers the implant itself in more depth.

Calibration: from fingersticks to the factory

A raw sensor current means nothing until software knows how much current equals how much glucose. That ratio is the sensor's sensitivity, and calibration sets it [8]. The first CGM, which MiniMed marketed in 1999, recorded glucose for three days, needed a fingerstick calibration every 6 to 12 hours, and showed nothing in real time; a healthcare provider downloaded the data afterward [10].

Factory calibration moves that step into manufacturing. It works only when sensors are stable during wear and consistent from one to the next, and it removes a hidden risk: a fingerstick calibration taken while something distorts the signal can leave every later reading off in the same direction [8]. As of October 2026, sensors handle calibration differently:

• Dexcom G7 15 Day: factory calibrated, with optional fingerstick calibration [1].
• FreeStyle Libre 3 Plus and 2 Plus: factory calibrated, and the user can't calibrate them [2].
• MiniMed Guardian 4: no calibration required when used with the MiniMed 780G pump, though the system uses any meter reading entered to calibrate. The older Guardian Sensor (3) needed a meter reading at least every 12 hours on the same pump [11].
• Eversense 365: still calibrated by fingerstick, with four calibrations within 36 hours after a 24-hour warm-up, then daily through day 13, then weekly [3].

The G6 sat between the two approaches. Entering the code printed on each sensor replaced fingerstick calibration for its 10 days, and users who skipped the code calibrated by fingerstick instead [5].

Warm-up and wear time

A new sensor needs time before its first reading. FDA lists 27 minutes for the 10-day Dexcom G7 and 60 minutes for the G7 15 Day [1]. The G6 needed two hours [5]. The Eversense 365 needs a full day [3].

Longer wear isn't guaranteed wear. Here's what surprised us in the G7 15 Day review: of 131 sensors analyzed, 73.9% lasted the full 15 days, and 25 (19.1%) shut off early when the sensor's own self-diagnostics ended the session [1].

Judging accuracy: MARD and the iCGM rules

MARD, short for mean absolute relative difference, averages how far each sensor reading lands from a laboratory reference value, as a percentage. Lower is better. The G7 15 Day's MARD ranged from 7.3% to 8.9% across its wear period, higher at the start and end than in the middle [1]. The Eversense 365 stayed between 7.8% and 9.9% in every 30-day interval FDA reported across its year [3]. Those numbers come from separate studies with different participants, so compare them with care; our comparative accuracy and MARD analysis goes further.

An average can hide bad misses, so FDA's iCGM regulation adds limits of its own. Across the measuring range, the lower 95% confidence bound for adult readings within 20% of the lab value must exceed 87% [6]. It also bans outright reversals: no sensor reading under 70 mg/dL can pair with a blood value over 180 mg/dL, or the other way around [6].

What can throw a sensor off

Some substances push readings in the wrong direction, and the labeling names them. For the Dexcom G7 15 Day, readings run falsely high with more than a standard acetaminophen dose (above 1 gram every 6 hours in adults) and with hydroxyurea [1]. For the Eversense 365, tetracycline-class antibiotics can falsely lower readings, and intravenous mannitol or sorbitol can falsely raise them [3]. Whether a medication matters for you is a decision for your care team. Our CGM interference profiles comparison lines up the labeling across brands.

One pattern stood out to us across these FDA records. The glucose oxidase reaction FDA described for the G6 in 2018 is the same one it described for the G7 15 Day in 2025 [5][1]. What changed was everything around it: 10 days of wear became 15, a two-hour warm-up became 60 minutes, and a sensor code gave way to factory calibration [5][1].