Tear and Saliva Glucose Correlation Failures

Tear and Saliva Glucose Correlation Failures

Introduction

Tears and saliva look like the perfect way around needles, and researchers have measured tear glucose for about 80 years [3]. What stopped them was mostly physiology rather than electronics: glucose in these fluids does not follow blood glucose closely or reliably enough to stand in for it [2][3]. As of October 2026, no glucose monitor that reads tears or saliva appears in FDA's device classification or 510(k) databases [11][12]. Three problems come up again and again:

• Lag time: glucose has to leave the blood and reach the tear film before a sensor can see it. One review puts the delay for tears at about 10 minutes [2]. Saliva has its own timing problem: glucose in saliva from the parotid gland stays raised for at least 2 hours after a meal, in people with and without diabetes [4].
• Low glucose concentrations: tear glucose runs roughly a tenth of blood glucose, about 0.05 to 0.5 mM in healthy people [2]. Saliva is similarly low, about 0.23 to 0.38 mM in healthy people and 0.55 to 1.77 mM in people with diabetes [2]. (mM, or millimoles per liter, is the unit most of these studies use.) The university team behind one saliva sensor says glucose exists in saliva only in minute concentrations, so detecting it takes a very sensitive sensor [10].
• Collection effects: the way tears are collected changes the result. Here's what surprised us. Irritating the conjunctiva, the thin membrane lining the eyelid, gives the highest tear glucose readings, and methods that avoid stimulating tears give the lowest [3]. The whole tear film holds only about 6 to 8 µL and turns over at 1 to 2 µL per minute [8], so even a small change in flow shifts what a sensor sees.

Key Failures and Findings

• Verily/Alcon Smart Lens: in November 2018, Verily, the life-sciences company in Google's parent group Alphabet, said it was halting work on the glucose-sensing contact lens it had been developing with Alcon [1][5][6]. Reviews that cite the announcement describe two problems: tears carry very little glucose, and other molecules in the tear fluid interfered with the readings [6]. A 2025 review sums it up as tear-to-blood correlations that were variable or lagged, plus confounders from the tear film and the environment [7]. None of these sources blames tear film lipids specifically. More on lens designs is in our article on smart contact lenses and tear glucose.
• Saliva Issues: a 2017 systematic review of saliva glucose in type 1 diabetes found that the timing and method of collection affect the measured value, and that many of the included studies were of poor methodological quality [4]. The reviewers concluded that saliva glucose might help track blood glucose, but only well-designed prospective studies could show it [4]. Another review notes that the link between blood and saliva glucose exists but is not well understood, which makes it hard to tie a saliva reading to a treatment decision [2].
• Current Status (October 2026): two projects show how slowly the field moves.
• NovioSense built a flexible, spring-shaped coil about 15 mm long and 1.3 mm wide that sits under the lower eyelid [8]. In a 2018 study of six adults with type 1 diabetes, worn for up to about 5 hours, 95% of its readings fell in the A and B zones of a consensus error grid, a chart that sorts readings by how risky their errors would be. The authors noted the study was not designed to test the ISO 15197 accuracy standard [8]. A 24-patient follow-up, funded by the company, reported a mean absolute relative difference (MARD, the average gap from reference readings) of 16.7% after about 6 hours in the eye [9]. Its development status since then is unclear.
• The saliva biosensor associated with The iQ Group is a printed transistor strip, about the size of a stick of chewing gum, coated with the enzyme glucose oxidase. It was developed at the University of Newcastle in Australia, licensed to The iQ Group in 2016, and designed to replace fingerstick blood testing [10]. Its current status is unclear: we found no recent update from the university or the companies that licensed it.

Minimally invasive sensors have moved further. In September 2025, FDA granted De Novo authorization to the Biolinq Shine Autonomous Time-in-Range Microsensor, which FDA classified as a new device type, a "glucose range monitoring system" [13]. Our article on microneedle array sensors covers related designs.

Conclusion

Tears and saliva fail as blood-glucose stand-ins for the same few reasons: they carry little glucose, they respond late, and the act of collecting them changes the answer [2][3][4]. Small studies of the NovioSense coil, run with company researchers, have looked promising over a few hours of wear [8][9]. The open question is whether any tear or saliva sensor can hold that accuracy for days, in larger groups, through the blinking, eating and crying of everyday life.