Non-Invasive Glucose Sensing Technologies

Introduction to Non-Invasive Glucose Sensing Technologies

Non-invasive glucose sensing means estimating glucose without a lancet, a needle or a sensor wire under the skin. As of October 2026, no such device is a stand-in for a fingerstick meter or a CGM. The FDA says it has not authorized, cleared or approved any smartwatch or smart ring that measures or estimates blood glucose on its own [1]. The idea is old, though, and one device did win FDA approval back in 2001.

Non-invasive or minimally invasive?

A 2022 review that catalogued bloodless glucose products sorted them into noninvasive optical devices, noninvasive fluid-sampling devices and minimally invasive devices, and counted 28, 6 and 31 products in those groups [2]. An optical device sends light through intact skin. A fluid-sampling device collects a body fluid, such as interstitial fluid (the fluid that bathes the cells under the skin), sweat or tears. A minimally invasive device still enters the skin. It just goes less deep than a fingerstick.

By that definition, the CGMs people wear today are minimally invasive; the FDA itself describes CGMs as devices "that pierce the skin" [1]. Microneedle designs sit on the same side of the line, only shallower. The Biolinq Shine, which the FDA granted De Novo authorization in September 2025, uses seven tiny electrodes that rest in the papillary dermis, the layer just beneath the skin's surface, and stores data for up to five days [3]. FDA reviewers call it an intradermal sensor, and its label limits it to adults 22 and older who don't use insulin [3]. Our article on microneedle array sensor innovation covers that design in more depth.

How the main techniques work

A 2020 review in Sensors groups non-invasive methods into optical, microwave and electrochemical approaches [4]. Each one reads glucose indirectly, through some physical property that changes when glucose does.

Infrared spectroscopy

Near-infrared (NIR) spectroscopy uses light from about 680 to 2,500 nanometers; mid-infrared (MIR) runs from 2,500 to 25,000 nanometers [4]. The device measures how much light the tissue absorbs at wavelengths where glucose absorbs. Water is the catch. It has strong NIR absorption peaks near 1,350 to 1,520 nm and 1,790 to 2,000 nm, so designers work in the gaps between them [4].

Raman spectroscopy

A laser hits the tissue and a small share of the light scatters back at shifted frequencies that depend on each molecule's chemical bonds [4]. Raman bands overlap less than infrared absorption does, which helps tell glucose apart from its neighbors. The trade-off is strength: glucose gives off an extremely weak Raman signal that background noise can easily bury [4].

Radio-frequency and microwave sensing

These sensors read the tissue's dielectric properties, meaning how it stores and responds to an electric field, which shift as glucose changes [4]. Here's what surprised us: the same review says exactly where in the tissue those changes happen, and why, is still unclear [4]. The startup Know Labs says its sensor sweeps radio frequencies from 300 to 4,400 MHz and turns the voltage readings into glucose estimates with machine-learning algorithms it keeps as trade secrets [5]. That description comes from the company's own page. It is not an FDA finding.

Reverse iontophoresis

A small electric current pulls glucose out through intact skin into a gel pad, where an enzyme sensor measures it [4]. The GlucoWatch, covered below, worked this way.

Why the signal is so hard to read

Glucose makes up only about 0.07 to 0.1% of plasma, and water, pigments and proteins in tissue also absorb light and interfere with the reading [4]. For optical coherence tomography, another optical method, the review adds sensitivity to skin temperature and movement, plus calibration models that don't carry over reliably between people [4]. It also cautions that a model's accuracy may not hold across groups that differ in age, skin color, gender or skin disease [4].

Timing is a second problem. Glucose in interstitial fluid trails blood glucose by about 4 to 10 minutes [4]. Tears and sweat are harder still: sensors that measure them face poor correlation with blood glucose [4]. We cover that gap in tear and saliva glucose correlation failures.

The one that reached the market: GlucoWatch

The FDA approved the GlucoWatch Automatic Glucose Biographer on March 22, 2001, under premarket approval P990026 [6]. The label was narrow: adults 18 and older, for detecting trends and tracking patterns, and "adjunctive," meaning it supplemented standard home glucose meters rather than replacing them [6]. In August 2002 the FDA approved the GlucoWatch G2 Biographer and extended use to children and teens ages 7 to 17 [7]. One detail caught our eye: the FDA files the device under product code NCT, "Instrument, Glucose, Noninvasive Technology" [7], even though it worked by drawing fluid through the skin with an electric current.

Worn like a watch, it ran a current of about 300 µA through the skin, and hydrogel disks containing the enzyme glucose oxidase reacted with the glucose that came out [4]. It needed fingerstick calibration and a long warm-up, its gaskets had to be changed every 12 hours, it couldn't read during sweating, and prolonged current could damage the skin [4].

A 2005 randomized trial in 200 children and teens with type 1 diabetes showed how those limits played out. Over six months, adding the G2 Biographer to standard monitoring did not improve HbA1c (a lab measure of longer-term average glucose) or reduce severe low-glucose events [8]. Use fell from 2.1 to 1.5 times a week. Participants blamed skin irritation (76%), frequent skipped readings (56%), excessive alarms (47%) and inaccurate readings (33%) [8]. Sales stopped in 2007, according to the 2020 review [4], and the FDA record lists the approval as withdrawn on April 23, 2010 [6].

The accuracy bar a new device would face

A CGM built to send readings to automated insulin dosing systems falls under the FDA's integrated CGM (iCGM) special controls [9], and a non-invasive sensor aiming for that role would face them too. They don't set a cutoff for MARD (mean absolute relative difference, the average percentage gap between sensor and reference readings). They require set shares of readings to land within fixed ranges of a reference blood glucose value. Across the measuring range, for example, the lower one-sided 95% confidence bound for readings within ±20% of the reference must exceed 87% [9]. Our article on regulatory pathways for non-invasive devices walks through the other routes a new sensor could take.

What the 2024 FDA warning says

In February 2024 the FDA warned against smartwatches and smart rings that claim to measure glucose without piercing the skin. It said dozens of companies make them, sold under many brand names, and that they do not directly test blood glucose [1]. An inaccurate reading can lead to the wrong dose of insulin or a sulfonylurea, and too much of either can quickly cause dangerously low glucose [1]. Its advice is not to buy or use these devices, and to ask a health care provider about an FDA-authorized one [1]. Which device fits your diabetes care is a decision for you and your care team.

As of October 2026, that warning is still posted under its February 21, 2024 issue date [1]. What nobody can say yet is whether any optical or radio-frequency sensor can clear the iCGM bar across the full range of people who would wear it, every age and skin tone included.