Regulatory Pathways for Non-Invasive Devices

As of October 2026, the FDA has not authorized any glucose monitor that works without piercing the skin. In February 2024 the agency said it had not authorized, cleared or approved any smartwatch or smart ring that measures or estimates blood glucose on its own [1]. The only noninvasive glucose instrument it ever approved, the GlucoWatch Automatic Glucose Biographer, was approved in March 2001, and that approval was withdrawn on April 23, 2010 [2]. So the real question for anyone building one is which regulatory route it would take. When we read the FDA's records, the answer turned less on how a device senses glucose than on what its maker says it is for.

What counts as non-invasive

The FDA's 2024 warning covers any watch or ring that claims to read glucose without a finger prick or a needle, whatever the brand [1]. It separates those products from smartwatch apps that only display readings from authorized devices that do pierce the skin, such as continuous glucose monitors (CGMs) [1]. The agency's January 2026 general wellness guidance draws the line more precisely. In one example, a skin product counts as noninvasive because it doesn't penetrate the stratum corneum, the outermost layer of skin [3]. In another, a glucose wearable built on microneedles falls outside the low-risk wellness category for the opposite reason: it goes through that layer [3].

Three classes, three routes

The FDA sorts devices into classes by risk. A device's class depends on its intended use and its indications for use [4]. Each class points to a different premarket route:

• Premarket approval (PMA) is the scientific and regulatory review for Class III devices, the highest-risk category, where the FDA has decided general and special controls alone can't assure safety and effectiveness [5].
• 510(k) clearance means showing that a device is "substantially equivalent" to a legally marketed device, which the FDA calls the predicate [6].
• De Novo classification is for novel devices with no legally marketed predicate, when general controls, or general and special controls, give reasonable assurance of safety and effectiveness [7].

The GlucoWatch: one approval, then a withdrawal

The GlucoWatch went through PMA under product code NCT, "instrument, glucose, noninvasive technology" [2]. Its 2001 approval covered adults 18 and older. The label framed it narrowly, as a tool for detecting trends and tracking patterns that would supplement, not replace, standard home glucose meters [2]. A 2002 supplement approved the GlucoWatch G2 Biographer and extended the indications to children and teens aged 7 to 17 [8].

In 2004 the FDA approved G2 labeling that described pre-treatment products and a method to reduce skin irritation without affecting accuracy [9]. The FDA database lists Animas Corp. as the applicant and April 23, 2010 as the withdrawal date [2]. The record doesn't say why, and we haven't found a primary source that does.

Here's what surprised us: the NCT product code is still on the books. As of October 2026, the FDA's classification database lists it as Class 3, with PMA as the submission type [10]. A company filing under that code today would face the same highest-risk route the GlucoWatch did.

Why most CGMs sit in Class II

Integrated CGMs, or iCGMs, take a different road. Under 21 CFR 862.1355, an iCGM measures glucose in bodily fluids continuously or frequently and reliably sends readings to connected devices, including automated insulin dosing systems. The regulation makes it Class II with special controls [11]. That category came from a De Novo: on March 27, 2018, the FDA granted Dexcom's request for the Dexcom G6 [12]. Dexcom has since replaced the G6 with its G7 sensors and says it will no longer manufacture the G6 after July 1, 2026 [13].

The special controls are where the hard work lives. Makers must submit clinical data comparing sensor readings with a laboratory reference method, from a study that covers the intended users and the device's whole measuring range [11]. Performance has to hold up across the full wear period [11]. In adults, the lower 95 percent confidence bound for readings within 20 percent of the reference must exceed 87 percent across that range [11]. The rule also requires interference testing and a usability study [11].

The regulation's identification text says "bodily fluids" and names no sensing method [11]. A non-invasive sensor that wanted this Class II route would still have to clear the same accuracy bar as a filament under the skin.

A new category for a minimally invasive sensor

A recent De Novo shows how the FDA handles a novel glucose sensor. On September 23, 2025, the FDA granted Biolinq's request for the Biolinq Shine Autonomous Time-in-Range Microsensor and created a Class II device type, the "glucose range monitoring system," under 21 CFR 862.1359 [14]. It isn't non-invasive. Its seven microelectrodes sit in the papillary dermis, the layer just under the skin's surface, and measure glucose in interstitial fluid [15]. The labeling in that decision was prescription-only, for adults 22 and older who don't use insulin, and the sensor shows colored lights instead of numbers [15]. Our article on microneedle array sensors covers the technology.

The 2026 wellness opening, and where it closes

The January 6, 2026 wellness guidance, which replaced a 2019 version, says the FDA may treat products that use non-invasive sensing, such as optical sensing, to estimate blood glucose as general wellness products when the output is meant solely for wellness [3]. The conditions are strict. Such a product can't be intended to diagnose or treat a disease, can't substitute for an FDA-authorized device, and can't prompt specific clinical action; any values that mimic clinical ones have to be validated [3].

The same guidance says glucose meters, CGMs and any product whose outputs feed treatment decisions such as medication dosing are not wellness products [3]. Labeling that names a disease, cites a diagnostic threshold or claims medical-grade accuracy also takes a product out of the category [3]. When we checked in October 2026, the 2024 smartwatch warning was still posted without an update [1].

After authorization

Getting to market doesn't end the FDA's oversight. Manufacturers must report when one of their devices may have caused or contributed to a death or serious injury. They must also report malfunctions that would be likely to do that if they recurred [16].

The accuracy problem behind the rules

Researchers have chased non-invasive glucose measurement for decades because the finger prick is painful, inconvenient and carries a risk of infection [17]. The obstacle is signal quality: one review notes that patient motion and sweat at the skin surface interfere with electrical (dielectric) glucose measurements [17]. The FDA's worry is concrete. Inaccurate readings can lead to the wrong dose of insulin, sulfonylureas or other drugs that rapidly lower glucose [1]. Which monitor fits your treatment is a decision to make with your care team. For the sensing methods still in research, see our overview of non-invasive and experimental [glucose monitoring](/continuous-glucose-monitoring/non-invasive-and-experimental-glucose-monitoring/).