Tear and Saliva Glucose Monitoring

Tear and Saliva [Glucose Monitoring](/)

Tears and saliva both carry glucose, and researchers have spent decades trying to read it without a needle. None of that work has reached the market. As of October 2026, we searched FDA's 510(k), premarket approval and device classification records for a tear or saliva glucose monitor and found no matches [1][2][3]. What exists instead is a trail of prototypes and small human studies, plus two 2026 reviews that explain why tears look more promising than saliva.

Why these fluids are hard to measure

The first problem is how little glucose there is to find. When a University of Washington team built its 2011 contact lens sensor, it designed it for tear glucose of 0.1 to 0.6 millimoles per liter (mM), which works out to roughly 2 to 11 mg/dL [4]. Saliva holds even less. A 2020 mouthguard study put typical salivary glucose at 20 to 200 micromoles per liter, or 0.02 to 0.2 mM [5].

Collection is the second problem. A 2007 review of 80 years of tear studies found that the sampling method drove the answer: studies that irritated the conjunctiva (the membrane lining the eyelid) measured the highest tear glucose, and studies that avoided stimulating tears measured the lowest [6]. The same review reported that researchers still disagreed about whether tear glucose tracks blood glucose at all [6].

Saliva adds chemical interference. The Tokyo mouthguard team reported that ascorbic acid (vitamin C) and uric acid in saliva hinder accurate glucose measurement, so it coated its electrode with a cellulose acetate membrane to screen them out [5]. We cover the physiology in more depth in Tear and Saliva Glucose Correlation Failures.

Tear glucose devices, 2004 to 2020

Color-changing and current-measuring lenses (2004 and 2011)

Early tear work was chemistry on a bench. In 2004, University of Pittsburgh chemists described a photonic crystal material: a hydrogel holding an ordered array of tiny particles, whose diffracted color shifted from red to blue as glucose rose across the range found in tears. They proposed it for eye inserts or diagnostic contact lenses [7].

The 2011 University of Washington lens took the electrochemical route. Its amperometric sensor measures the electric current produced when the enzyme glucose oxidase reacts with glucose. The team added a Nafion layer to reduce interference from ascorbic acid, lactate and urea in tears, and reported a response time of 20 seconds in testing [4].

Google, Novartis and the smart lens (2014 to 2018)

The best-known attempt began on January 16, 2014, when Google[x] said it was testing a soft contact lens with a tiny wireless chip and a miniaturized glucose sensor sealed between two layers of lens material. Its prototypes could take a reading once per second, and Google said it was in discussions with the FDA and would look for partners to bring the lens to market [8]. Google and Novartis jointly announced the project that same year [9]. The work ended up as a joint project of Alcon, then a Novartis unit, and Verily, Google's life-sciences company [10].

It did not last. A 2021 review of CGM history reports that the project was abandoned in November 2018 because measuring glucose in tears proved unreliable [10]. Here's what surprised us: a 2022 review argues the program's stated reason was narrower than "tears don't follow blood." It points to the small amount of glucose in tears and to contaminants that interfered with the reading [9]. Our Smart Contact Lenses & Tear Glucose article covers lens designs in detail.

NovioSense: a sensor under the eyelid (2018 to 2020)

NovioSense, a Dutch company, skipped the lens. Its sensor sits under the lower eyelid, in the pocket called the inferior conjunctival fornix, and reads glucose in basal tears, the steady tear film that isn't triggered by irritation [11]. In a 2018 phase II trial, six people with type 1 diabetes wore it, and the authors reported that its readings correlated well with fingerstick blood glucose on a consensus error grid, a chart that sorts readings by how risky their errors would be [11].

A follow-up published in October 2020 added 24 more people with type 1 diabetes. After about 6 hours in the eye, the sensor's mean absolute relative difference (MARD, the average percentage gap between sensor and reference readings) was 16.7%, and its median was 13.3%. The team converted the raw tear signal into blood values with a neural-network calibration model [12]. Both papers list NovioSense staff among their authors, and NovioSense paid for the second study [11][12].

Saliva glucose devices, 2016 to 2020

Saliva work has centered on the mouth itself. In 2016, a Tokyo Medical and Dental University team built a mouthguard with platinum and silver/silver chloride electrodes, a glucose oxidase membrane and a wireless measurement system. In artificial saliva it detected glucose from 5 to 1,000 micromoles per liter, and it ran for more than 5 hours in a model jaw built to imitate the mouth [13]. The authors framed it as a tool for dental patients [13].

That same year, a French group reported an array of organic electrochemical transistors (OECTs), small transistors made from carbon-based materials, fed by a "finger-powered" microfluidic channel. It measured glucose, lactate and cholesterol, was checked with human samples, and was aimed at low-concentration fluids such as saliva [14].

The Tokyo team came back in 2020 with the cellulose acetate-coated mouthguard described above. It measured glucose in saliva without any pretreatment, and its output tracked concentration in samples from healthy volunteers [5]. As their abstracts describe them, none of these three saliva studies tested a device in people with diabetes.

What the 2026 reviews found

The tear evidence looks better on paper. A 2026 systematic review examined 12 smart contact lens studies in people and animals. Tear glucose correlated strongly with blood glucose, MARD values ranged from 12.5% to 16.7%, and tear glucose trailed blood glucose by 5 to 10 minutes. The reviewers still called for validation in larger, more diverse groups under real-world conditions [15]. NovioSense's 16.7% sits at the top of that MARD range [12][15].

Saliva fared worse. A February 2026 meta-analysis of 25 studies compared saliva and blood samples taken after at least 8 hours of fasting. Whole-mouth saliva explained only about 5% of the variation in blood glucose, and saliva from the parotid gland (the large salivary gland in front of the ear) about 11%. The authors concluded that current evidence doesn't support saliva glucose as a replacement for blood glucose measurement, at least for now [16].

Where things stand in October 2026

No tear or saliva glucose monitor appears in FDA's records [1][2][3], so every reading described above came from a research device under study conditions. Whether any glucose reading is reliable enough to act on is a decision for your care team. The open question is the one the 2026 tear review raised. Tear sensors have done well in small studies, some lasting only hours, and nobody has yet shown that accuracy holding up in larger, more diverse groups under real-world conditions [12][15].