Skip to content
type1.science

Non-invasive glucose monitoring (Apple, Samsung, Know Labs, Afon Glucowear)

Various (Apple, Samsung, Google, Know Labs, Afon Technology, RSP Systems)

The long-promised needle-free CGM — read glucose through intact skin with light (optical / Raman / near-infrared) or radio waves (RF / Bio-RFID), no sensor filament under the skin. As of 2026 it remains research, not a product: no medical-grade, commercially available non-invasive CGM exists, the FDA warns against any smartwatch or ring that claims to measure glucose without piercing the skin, and the best peer-reviewed device accuracy is still calibration- dependent and tested mostly in type 2 diabetes — not yet reliable enough to dose insulin from.

Years awayEarly evidencecgmnon-invasive

The scorecard

Accuracy30

No needle-free system is accurate enough to dose insulin from. The best peer-reviewed dedicated device (RSP Systems, Raman) reached a 12.8% MARD in 50 people with type 2 diabetes over 2 days — worse than the ~8% of modern minimally invasive CGMs, and only after individual calibration. Know Labs' RF sensor reports ~11–14% MARD by range; consumer-watch optical methods are far less validated. Honest score: promising signal, not clinical-grade.[3]

Low lag55

Optical/RF methods read at or near the skin surface and could in principle track blood glucose with less physiological delay than interstitial sensors, but no needle-free device has demonstrated a real-world lag advantage in T1D; scored as unproven potential, not a delivered benefit.[2]

Interoperability10

Nothing to integrate — no non-invasive system is approved as an iCGM, so none can legally drive a pump, AID system or DIY loop, and the FDA explicitly warns against using any glucose-claiming smartwatch/ring for dosing.[1]

Sensor lifespan90

The one clear win of the concept: a needle-free optical/RF reader is part of a watch, band or handheld with no disposable sensor to replace every 10–15 days — effectively unlimited wear once the hardware works.[9]

No calibration25

Today's needle-free methods still need per-person calibration against fingersticks or venous blood; the leading Raman device individualizes a pre-trained model with 10 measurements, and factory calibration remains a stated goal, not a reality. Higher score means less calibration — so this is low.[3]

Alerts & prediction20

Continuous, validated high/low and predictive alerts depend on a continuous, accurate signal that no non-invasive system reliably delivers yet; demos are spot or short-session, so trustworthy alerting is not established.[1]

Form factor85

The whole appeal — no filament under the skin, no adhesive patch, glucose read from a watch, wrist band, or finger; Afon builds it into a watch strap and Apple's target is the Apple Watch. Form factor is the strength, accuracy the gap.[6]

Ketone sensing0

No non-invasive glucose program publicly offers continuous ketone sensing; these efforts are focused solely on cracking glucose first.

Access & cost5

You cannot buy a medical-grade non-invasive CGM anywhere in 2026. Watches and rings that claim glucose are unproven and FDA-warned; dedicated devices are in trials. Access is essentially zero, hence the floor score.[1]

Editor’s take

We rank non-invasive glucose monitoring as the most over-hyped and most watched item in the category. The dream is real and the physics is finally yielding — Raman devices now post low-double-digit MARDs and Apple has reached proof-of-concept — but in 2026 there is still no needle-free CGM you can dose insulin from, and the FDA's standing warning against glucose-claiming smartwatches and rings is the line we hold. We list it as research/early so people with T1D can track the genuine progress without betting their safety on a device that does not yet exist. The first credible win will likely be a trend/wellness reader, with insulin-dosing accuracy years behind it.

The full picture

Non-invasive glucose monitoring is the long-promised needle-free CGM: read your glucose straight through intact skin — with light or radio waves — so there is no tiny sensor filament under the skin and no adhesive patch. It is the single most hyped idea in diabetes tech. The honest 2026 headline: it does not exist as a medical product yet. No needle-free, medical-grade, commercially available CGM is on the market, and the US FDA has issued a standing warning that it has not authorized, cleared, or approved any smartwatch or smart ring that claims to measure blood glucose without piercing the skin — because a wrong reading can lead to a wrong insulin dose, a dangerous low, coma, or death.1

How it's supposed to work. Two main families. Optical methods shine specific wavelengths of light into the skin and read what comes back: Raman spectroscopy (Apple's chief rival approaches and Samsung's research), near-infrared / optical absorption spectroscopy (Apple's silicon-photonics "Project E5"), and others like photoacoustic and terahertz.23 Radio-frequency methods (Know Labs' Bio-RFID and Afon's Glucowear) bounce radio waves through tissue and read how glucose changes the signal.45 The hard part is the same for all of them: the glucose signal is tiny and buried under skin, water, and movement.3

Accuracy — the whole problem. A modern minimally invasive CGM has a MARD (mean absolute relative difference; lower is better) around 8%. The best peer-reviewed non-invasive device — a Raman system tested in 50 people with type 2 diabetes — reached 12.8% MARD over two days, and only after individual calibration, with 100% of readings in the safe zones of an error grid.6 Know Labs' RF sensor reports roughly 11–14% MARD depending on glucose range, against venous blood, also in type 2 diabetes/prediabetes — not yet FDA cleared.4 Afon's early Glucowear data was cruder still: in one person it correctly sorted about 74% of readings as "normal vs elevated."5 The foundational science is real — MIT and Samsung published the first direct observation of the glucose Raman fingerprint in living skin in 2020 — but that was in pigs, and it is a long road from there to dosing insulin in a person with T1D.7

Wear, warm-up, calibration. The appeal is obvious: a watch, band, or finger reader means no disposable sensor to change every 10–15 days. But today's methods still need per-person calibration against fingersticks or blood, and "factory calibration" is a stated goal, not a reality.6

What it works with. Nothing, for dosing. Because no non-invasive system is an approved iCGM, none can legally drive a pump, AID system, or DIY loop, and the FDA warns against trusting any glucose-claiming watch or ring for treatment decisions.1 There is no ketone sensing in any of these programs.

Who it's for, access & cost. As of 2026, effectively no one for clinical use — there is nothing medical-grade to buy in the US, UK, EU, Canada, Australia, Japan, or elsewhere. Dedicated devices are in trials.65

What's coming. Apple's optical sensor is at proof-of-concept, with outside reporting estimating 2028–2030+ before any Apple Watch product.2 Samsung is openly working on an optical Galaxy Watch monitor with no date. Know Labs is building toward an FDA submission, and Afon is presenting fresh accuracy data at the ATTD 2026 congress.45 Newer work is shrinking Raman hardware toward something wearable.8 The likeliest first win is a trend/wellness reader; an insulin-dosing-grade non-invasive CGM for T1D is still years out.

References

  1. U.S. Food & Drug Administration. Do Not Use Smartwatches or Smart Rings to Measure Blood Glucose Levels: FDA Safety Communication (21 Feb 2024). https://www.fda.gov/medical-devices/safety-communications/do-not-use-smartwatches-or-smart-rings-measure-blood-glucose-levels-fda-safety-communication ↩ ↩2

  2. BGR (reporting Bloomberg / Mark Gurman). Apple's Made Major Progress With Future Blood Glucose Monitor For Watch (2023) — Project E5: silicon photonics + optical absorption spectroscopy, proof-of-concept stage, years from market. https://www.bgr.com/tech/apples-made-major-progress-with-future-blood-glucose-monitor-for-watch/ ↩ ↩2

  3. Alsunaidi B, et al. A Review of Non-Invasive Optical Systems for Continuous Blood Glucose Monitoring. Sensors 21:6820 (2021). PMID 34696033. https://pmc.ncbi.nlm.nih.gov/articles/PMC8537963/ ↩ ↩2

  4. Know Labs. Clinical Research & Development (Bio-RFID / RF dielectric sensor) — reported MARD ~12.7% normoglycemic / 14.0% hyperglycemic, and ~11.1% overall vs venous blood in prediabetes/type 2 diabetes; progressing toward FDA submission, not yet cleared. https://www.knowlabs.co/research-and-development ↩ ↩2 ↩3

  5. Afon Technology / Drug Delivery Business News. Afon highlights accuracy of its Glucowear non-invasive glucose monitor — RF/microwave wrist wearable; preliminary single-subject evaluation correctly classified ~74% of readings (sensitivity 87%, specificity 85%); in development, no CE/FDA clearance; data slated for ATTD 2026. https://www.drugdeliverybusiness.com/afon-highlights-accuracy-non-invasive-cgm/ ↩ ↩2 ↩3 ↩4

  6. Pors A, et al. (RSP Systems & Institute for Diabetes Technology, Ulm). Calibration and performance of a Raman-based device for non-invasive glucose monitoring in type 2 diabetes. Scientific Reports 15:10226 (2025) — n=50, MARD 12.8% (95% CI 12.4–13.2), 100% in consensus error grid zones A+B after individual calibration. PMID 40133405. DOI 10.1038/s41598-025-95334-x. https://pmc.ncbi.nlm.nih.gov/articles/PMC11937273/ ↩ ↩2 ↩3

  7. Kang JW, et al. (MIT & Samsung Advanced Institute of Technology). Direct observation of glucose fingerprint using in vivo Raman spectroscopy. Science Advances 6:eaay5206 (2020). PMID 32042901. DOI 10.1126/sciadv.aay5206. https://pmc.ncbi.nlm.nih.gov/articles/PMC6981082/ ↩

  8. Bresci A, et al. (MIT & Apollon Inc.). Band-Pass Raman Spectroscopy Unlocks Compact Point-of-Care Noninvasive Continuous Glucose Monitoring. Analytical Chemistry 97:27020-27026 (2025). PMID 41342332. DOI 10.1021/acs.analchem.5c01146. https://pmc.ncbi.nlm.nih.gov/articles/PMC12713608/ ↩

Coming soon

ETA · No medical-grade product as of 2026. Dedicated devices in trials; Apple Watch glucose estimated 2028–2030+ by external reporting.

  • →Apple Project E5 — silicon-photonics optical sensor for the Apple Watch is at proof-of-concept; external reporting estimates 2028–2030+ before any product.
  • →Samsung — publicly working on an optical (Raman + near-infrared) non-invasive monitor for Galaxy Watch, no committed launch date.
  • →Know Labs (Bio-RFID / RF dielectric) — building its clinical dataset toward an eventual FDA submission; not yet cleared.
  • →Afon Glucowear (RF/microwave wrist wearable) — continuing accuracy and reliability studies; data slated for the ATTD 2026 congress.
  • →Raman device makers (e.g. RSP Systems) — pushing from individual calibration toward factory calibration, and from type 2 diabetes validation toward the tighter accuracy T1D insulin dosing requires.

Sources

  1. [1]Do Not Use Smartwatches or Smart Rings to Measure Blood Glucose Levels: FDA Safety Communication · regulatory · 2024-02-21 — FDA states it has not authorized, cleared or approved any smartwatch or smart ring intended to measure or estimate blood glucose without piercing the skin; inaccurate readings can cause wrong insulin/medication dosing, dangerous lows, coma or death. Retrieved and confirmed (HTTP 200).
  2. [2]Direct observation of glucose fingerprint using in vivo Raman spectroscopy · peer-reviewed · 2020-01-24 — Kang JW, et al. (MIT Laser Biomedical Research Center + Samsung Advanced Institute of Technology). Science Advances 6:eaay5206. First direct in vivo observation of the glucose Raman peak in skin (live swine clamp). Foundational science, not a product. PMID 32042901, DOI 10.1126/sciadv.aay5206. Via PubMed.
  3. [3]Calibration and performance of a Raman-based device for non-invasive glucose monitoring in type 2 diabetes · peer-reviewed · 2025-03-25 — Pors A, et al. (RSP Systems + Institute for Diabetes Technology, Ulm). Sci Rep 15:10226. n=50 type 2 diabetes, 2-day study, 1918 paired points: MARD 12.8% (95% CI 12.4–13.2), 100% in consensus error grid zones A+B, after a brief individual calibration. PMID 40133405, DOI 10.1038/s41598-025-95334-x. Via PubMed.
  4. [4]Band-Pass Raman Spectroscopy Unlocks Compact Point-of-Care Noninvasive Continuous Glucose Monitoring · peer-reviewed · 2025-12-04 — Bresci A, et al. (MIT + Apollon Inc.). Anal Chem 97:27020-27026. Band-pass Raman at 830 nm validated on tissue phantoms and human skin; a step toward shrinking Raman CGM out of the lab. PMID 41342332, DOI 10.1021/acs.analchem.5c01146. Via PubMed.
  5. [5]A Review of Non-Invasive Optical Systems for Continuous Blood Glucose Monitoring · peer-reviewed · 2021-10-14 — Alsunaidi B, et al. Sensors 21:6820. Survey of optical approaches (NIR, Raman, photoacoustic, terahertz) and why background tissue signal makes the glucose signal hard to isolate. PMID 34696033, DOI 10.3390/s21206820. Via PubMed.
  6. [6]Apple's Made Major Progress With Future Blood Glucose Monitor For Watch · news · 2023-02-23 — Reports Bloomberg/Mark Gurman's account of Apple's Project E5: silicon photonics + optical absorption spectroscopy, reached proof-of-concept, iPhone-sized prototype worn on the bicep, years from market. Retrieved (HTTP 200).
  7. [7]Clinical Research & Development — Know Labs (Bio-RFID / RF dielectric sensor) · manufacturer — Know Labs' R&D summary: RF dielectric sensor with AI model; reported MARD 12.7% (normoglycemic) / 14.0% (hyperglycemic) in a peer-reviewed analysis, and ~11.1% in an earlier study vs venous blood in prediabetes/type 2 diabetes; states it is progressing toward an FDA submission, with no clearance yet. Retrieved (HTTP 200).
  8. [8]Know Labs' Non-Invasive Glucose Monitor Achieves 11.1% MARD in Latest Clinical Research Study · manufacturer · 2024-03-06 — Know Labs press release reporting an overall 11.1% MARD for its RF dielectric sensor vs venous reference in prediabetes/type 2 diabetes. Press release, not peer-reviewed; corroborated by the knowlabs.co R&D page above.
  9. [9]Afon highlights accuracy of its Glucowear non-invasive glucose monitor · news — Afon Technology's Glucowear: RF/microwave wrist wearable in a watch strap; preliminary single-subject real-world evaluation correctly classified ~74% of readings as normal vs elevated (sensitivity 87%, specificity 85%); in development, no CE mark or FDA clearance reported; data slated for ATTD 2026. Retrieved (HTTP 200).
  10. [10]Glucowear™ — Afon Technology · manufacturer — Manufacturer page: needle-free wrist sensor using RF/microwave technology, streams to a smartphone app; described as in development, initially aimed at non-insulin-dependent type 2 diabetes, with no commercial date. Retrieved (HTTP 200).