Implantable Fluorescence Sensors

Implantable Fluorescence Sensors for Continuous [Glucose Monitoring](/)

An implantable fluorescence sensor reads glucose with light: its fluorescent indicator glows brighter when glucose binds, and an optical system inside the implant measures the glow [1]. The example on the market as of October 2026 is Senseonics' Eversense 365, which FDA cleared in September 2024 for up to 1 year of wear [1]. Here we stay with the chemistry and optics; insertion, calibration schedules and coverage are in our article on implantable long-term CGM solutions.

The indicator: boronic acids that grab glucose

The working part of the Eversense sensor is a glucose-indicating polymer, a hydrogel formed onto the surface of the sensor's housing. It is abiotic: there is no enzyme and no electrode [2].

A glucose molecule has pairs of neighboring hydroxyl groups, called cis-diols. The indicator has two boronic acid groups spaced to form a glucose-sized cleft, and the diols on glucose condense reversibly with those boronates [2]. FDA's summary of the 365 calls this "selective, fully reversible binding" to a proprietary indicator built from a monomer it names TFM [1].

The light comes from anthracene groups in the indicator. With no glucose bound, nearby amine groups quench their fluorescence through a process called photo-induced electron transfer, where an electron hop steals the energy that would otherwise leave as light. When glucose binds at the boronic acids, that quenching is disrupted and fluorescence rises, without chemically changing the indicator [2]. More glucose means more binding and a brighter signal.

The optical readout

Inside the sensor, a light-emitting diode (LED) excites the indicator, and two filtered photodiodes measure how strongly it fluoresces [2]. A 2014 Senseonics paper describes the package as a miniaturized, polymer-encased optical detection system coated with the indicator [3]. The finished sensor is about 3.5 mm by 18.3 mm and has no battery. The transmitter on the skin above it powers it by radio frequency, converts its signal to a glucose value, and sends that value to a phone app over Bluetooth Low Energy [1].

Reading glucose with light has its own blind spots. A substance can raise the signal by binding to the hydrogel or by emitting light in the sensor's spectral range, and lower it by absorbing the excitation light or quenching the indicator's glow [4]. In bench tests of 41 substances, only tetracycline and mannitol caused bias beyond ISO 15197:2013 limits at therapeutic concentrations. Tetracycline was tested because it absorbs or fluoresces at wavelengths that overlap anthracene's; mannitol is a sugar alcohol with diol groups like glucose's. Acetaminophen and ascorbic acid (vitamin C), which interfere with some enzymatic sensors, stayed within limits [2]. Whether any of this matters for a given person is a question for their care team.

Why the signal fades: oxidation

Because the binding is reversible and leaves the indicator intact, you might expect the chemistry to last indefinitely. Here is what surprised us: an early implanted version lost its signal within hours [5]. Post-explant analysis showed the boronate recognition element had been oxidized off the indicator, and the researchers traced it to reactive oxygen species, especially hydrogen peroxide, from the body's normal inflammatory response to the implant [5]. The PROMISE authors call this de-boronation, and it impaired glucose binding [4].

The first fix was a 3-nanometer layer of platinum sputtered onto the porous hydrogel surface, which in the 2013 paper extended useful life from 1 day to an expected 6 months or more [5]. The PROMISE investigators later said the platinum extended the original sensors' lifetime to up to 3 months [4].

The second fix was a decoy. In PROMISE, 43 participants wore a modified sensor whose hydrogel included 4-vinylphenylboronic acid, a sacrificial boronic acid (SBA). It serves as a target for reactive oxygen species and so protects the boronic acids that bind glucose [4]. Over 180 days, 90% of the SBA sensors survived, compared with 65% of the standard sensors, and their overall MARD (mean absolute relative difference from lab glucose readings) was 8.5%, against 9.1% [4].

Oxidation also explains why these sensors still ask for fingersticks. The PROMISE calibration algorithm tracks the change in sensitivity caused by indicator oxidation, and its authors note that factory calibration might suit a 2-week sensor but not one lasting 180 days [4].

The steroid collar

The other defense works on the inflammation itself. The Eversense 365 has a silicone collar that slowly releases 1.75 mg of dexamethasone acetate, an anti-inflammatory steroid, to minimize the foreign body reaction around the sensor [1]. In its 2024 De Novo review of the Eversense AP, FDA said that release reduces tissue inflammation to keep the sensor working over 6 months [6]. Since the damaging peroxide comes from inflammation [5], the steroid and the anti-oxidation chemistry attack one problem from two ends, though none of our sources measured how much the steroid alone cuts oxidation.

FDA's summary of the 365 doesn't say whether it uses the platinum layer or the sacrificial boronic acid [1], so we can't confirm which protections today's sensor carries. What we can report is performance: in the ENHANCE study, the 365 had an overall MARD of 8.8% through a year of wear with mostly one calibration per week, and 90% of sensors survived to day 365 [7].

Other fluorescence approaches in research

Boronic acids are only one way to make glucose produce light. A 2005 review lists other recognition molecules, including the plant lectin concanavalin A, enzymes such as hexokinase, and a bacterial glucose-binding protein. Many designs use FRET (fluorescence resonance energy transfer) between a donor and an acceptor dye, and some measure fluorescence lifetime, how long the glow persists, rather than brightness [8]. Our article on optical and fluorescence-based glucose transduction goes further into those methods.

The FiberSense system from EyeSense put a FRET assay, in which glucose displaces labeled dextran from concanavalin A, inside a polymer optical fiber that crossed the skin [9]. In six people with diabetes over up to 14 days, MARD was 8.3% in the upper arm and 11.4% in the abdomen [9]. In mice, University of Tokyo researchers implanted glucose-responsive fluorescent hydrogel fibers in the ear; fibers made with polyethylene glycol caused less inflammation and tracked blood glucose through the skin for up to 140 days [10].

Where the Eversense design is heading

Senseonics' next steps change the power source rather than the chemistry. Its Gemini sensor adds an integrated 1-year battery, so it could work as a real-time CGM with a transmitter or give flash readings without one. A later design, Freedom, would drop the transmitter and talk to a handheld device by Bluetooth. Neither is approved [11]. As of March 2026, Senseonics had FDA approval for an investigational device exemption, had enrolled the first patients in a Gemini pivotal trial, and expected to complete it in the second half of 2026 [12].