Non-Invasive Glucose Monitoring Technologies

Introduction to Non-Invasive [Glucose Monitoring](/) Technologies

Non-invasive glucose monitoring (NIGM) means estimating glucose without breaking the skin: no fingerstick and no sensor filament under the skin. Developers use three broad approaches. Optical Spectroscopy (NIR/Raman) shines near-infrared, mid-infrared or Raman laser light into the skin. RF Sensing reads how tissue responds to radio waves. Transdermal Extraction pulls glucose out through intact skin into a patch. As of October 2026, the FDA has not authorized any smartwatch or smart ring that measures or estimates blood glucose on its own, and it tells the public not to buy or use them [1].

Technical Background

Picking up a glucose-related signal through skin is the easy part. Separating it from everything else is hard. 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 measurement [9]. A 2025 workshop report on NIGM puts the core limit in physics terms: devices can detect signals that correlate with glucose, but isolating glucose from other biological variables in a wearable used in real life remains unsolved [2]. Large consumer-electronics companies have worked on the problem too: Samsung researchers co-authored a 2020 study with MIT that observed glucose's Raman peaks directly in the skin of three live pigs [8].

You'll often see a MARD under 10% quoted as the accuracy needed for insulin dosing. That isn't the FDA's test. MARD (mean absolute relative difference) is the average percentage gap between a sensor's readings and matched reference blood glucose values. The FDA's special controls for integrated CGMs (iCGMs), the class designed to send readings to connected devices such as automated insulin dosing systems, set agreement targets instead [4]. In adults, the lower 95% confidence bound for the share of readings within ±20% of the reference has to exceed 87% across the measuring range. Tighter ±15% (or ±15 mg/dL) targets and wider ±40% (or ±40 mg/dL) targets are checked separately in three glucose bands (below 70 mg/dL, 70 to 180 mg/dL and above 180 mg/dL) [4]. The rule also bars any iCGM reading below 70 mg/dL when blood glucose is above 180 mg/dL, and the reverse [4]. Workshop speakers added a warning about the number itself: MARD figures from early studies that use fingerstick meters as the reference can mislead when set beside validated clinical systems [2].

Three Companies and Their Approaches

Three companies show how differently these approaches work in practice. As of the November 2025 workshop, no non-invasive device had FDA clearance for direct glucose measurement [2], and since then the corporate picture has changed for two of the three:

• Know Labs: Know Labs built a radio-frequency sensor it called Bio-RFID. It uses dielectric spectroscopy, sweeping frequencies from 300 to 4,400 MHz through the tissue and turning the measured voltages into glucose estimates with machine-learning models [3]. In a 2024 study of 31 adults with prediabetes or type 2 diabetes, a model trained on the sensor's data sorted glucose into one of three classes (high, normal, low) correctly 93.37% of the time [5]. That is a three-way classification, not a numeric reading. The public company renamed itself USBC, Inc. in August 2025 [16], and on March 27, 2026 it sold the sensor business to an entity controlled by Know Labs' former chairman and CEO [10].
• DiaMonTech: DiaMonTech aims a mid-infrared quantum cascade laser at the skin at wavelengths between 8 and 12 µm, where glucose absorbs light in a pattern specific to the molecule. The absorbed light warms the skin very slightly, and the device picks up that heat at the surface, a method called photothermal detection [6]. In a 36-person clinical study of a tabletop prototype, the best of four algorithms reached a MARD of 20.7% and 19.6% in two sessions. The authors compare that with early CGMs the FDA cleared for adjunctive use, meaning alongside fingersticks [6]. At the November 2025 workshop, the company described a smartphone-sized handheld, D-Pocket, as a prototype with market entry planned for mid-2026 [2].
• Nemaura: Nemaura's sugarBEAT uses reverse iontophoresis to draw glucose from interstitial fluid out through the skin [7]. It is technically non-invasive because nothing pierces the skin. You still wear a disposable adhesive patch with a sensor, plus a rechargeable transmitter that sends a reading every 5 minutes to a phone app [15], and calibration uses a fingerstick [7]. sugarBEAT received a CE mark in Europe in May 2019 after a clinical program in 75 people with type 1 or type 2 diabetes [11], and Nemaura filed a premarket approval (PMA) application with the FDA in July 2020 [14]. In January 2024 Nemaura accepted Nasdaq's decision to delist its stock and said it would move to over-the-counter trading [12], and sugarBEAT's current status is unclear. The approach itself has FDA history: the FDA approved the reverse-iontophoresis GlucoWatch Automatic Glucose Biographer in March 2001 for tracking glucose trends alongside standard home meters, and that approval was withdrawn in April 2010 [13] [7].

Devices That Reached, or Nearly Reached, the Market

A few non-invasive monitors got further than a prototype, and their stories explain much of the caution around new claims. Several carried a CE mark in Europe, but a 2022 review found they were withdrawn or never sold [17].

• Pendra: This impedance-spectroscopy monitor received a CE mark in 2003. It was withdrawn because its readings were considered dangerously inaccurate, and its maker went out of business [17].
• C8 MediSensors: This San Jose company received a CE mark in 2012 for an optical glucose monitor based on Raman spectroscopy [23]. It couldn't raise the money to finish the design and shut down in 2013 without bringing the device to market [25].
• Google's glucose contact lens: Verily, the Alphabet company that ran the project, began it with Novartis in 2014 and halted it in 2018, after its trials found that tear glucose and blood glucose readings weren't consistent enough to meet the standards of a medical device [22]. The 2022 review reached a similar conclusion for other fluids collected without breaking the skin, such as sweat and saliva: their glucose tracks blood glucose poorly [17].
• GlucoWise: The UK company MediWise designed this device to read glucose by passing low-power millimeter-wave radio signals through the skin between the thumb and forefinger. As of the 2022 review, it had neither FDA clearance nor a CE mark [17], and we couldn't confirm its current status.
• Afon Technology: This Welsh company uses microwave spectroscopy [17]. Its website describes the goal as "a completely non-invasive, continuous blood glucose monitor" [21], and it listed a prototype study on ClinicalTrials.gov with a September 2021 start date [20].

Other groups look beyond the skin. Breath analysis with "electronic nose" devices looks for volatile organic compounds in exhaled breath that can carry markers of diabetes [24].

What Accuracy Claims Show So Far

Published accuracy data are scarce. Of 12 companies a 2021 review found announcing non-invasive glucose monitors, only one had published glucose data in a peer-reviewed medical journal [23]. Of the 28 non-invasive optical and radio-wave products in the 2022 review, seven reported a MARD, ranging from 5.7% to 21%. One more reported only a median error, and the other 20 gave no accuracy figure at all [17]. Those numbers were gathered from journals, company websites and news reports, so treat them as rough [17].

For comparison, CGMs that use a thin filament under the skin are tested against a lab analyzer in formal studies. The Dexcom G7 15 Day had an overall MARD of 8.0% in adults [18], and the 15-day sensor used in the FreeStyle Libre 2 Plus and Libre 3 Plus systems had 8.2% in adults and 8.1% in children ages 6 to 17 [19]. You can compare current sensors in our MARD analysis.

What Still Stands in the Way

Two problems stand between these devices and everyday diabetes care: the signal itself, and the evidence needed to trust it.

• Signal Interference: Water is the most abundant substance in body fluids and absorbs strongly in two near-infrared bands, so optical devices have to measure in the windows between those peaks [9]. Differences between people, including age and skin color, also cause large measurement errors [9]. A 2023 patent review adds hydration and some medications to that list [7]. Pigment matters most for visible light and for near-infrared wavelengths shorter than about 1 µm, while radio waves and microwaves aren't expected to be affected [17]. DiaMonTech argues that its infrared light reaches only about 100 µm into the skin, so pigmentation should matter little. Here's what surprised us: all 36 people in its 2025 study were white, so that claim hasn't yet been tested across skin tones [6].
• Signal Specificity: Glucose produces a weak spectroscopic signal, and a sensor has to pick it out from everything else in skin and tissue fluid, including albumin, urea, amino acids and vitamin C [17]. Indirect methods, which infer glucose from signals such as heart rate or blood flow, face a harder version of this problem because nothing in their signal comes from the glucose molecule itself [17].
• Regulatory Status: On February 21, 2024, the FDA warned against using smartwatches or smart rings that claim to measure blood glucose without piercing the skin. The agency said it has not authorized, cleared or approved any smartwatch or smart ring intended to measure or estimate blood glucose on its own. It also warned that inaccurate readings can lead to wrong doses of insulin or other medicines that lower glucose quickly [1]. The warning applies to every brand. It doesn't cover smartwatch apps that display data from an FDA-authorized CGM, which does pierce the skin [1]. Some developers propose selling non-invasive sensors as general wellness products that show trends without medical claims [2]. Nemaura took that route with proBEAT, a non-invasive glucose monitor and subscription service that it said would launch in the US as a general wellness product [14]. A wellness label is a regulatory category, not evidence that the readings are accurate enough to manage diabetes. Which device fits your care is a decision to make with your care team.

Where the Field Stands

The research keeps moving. Peer-reviewed studies in 2024 and 2025 reported RF classification and mid-infrared accuracy results, each in a few dozen adults [5] [6]. What's still missing is the kind of evidence the FDA asks of an iCGM: clinical data that represent the whole intended-use population and measuring range, with consistent performance over the entire sensor wear period [4]. We cover that process in regulatory pathways for non-invasive devices. Workshop participants agreed the field needs a shared benchmarking metric, which would make claims from different companies easier to compare [2]. Until a non-invasive device clears that bar, the FDA's advice for anyone whose care depends on accurate glucose readings is to ask a health care provider about an FDA-authorized device [1].