Study Shows Promise for Non-Invasive Blood Glucose Wearables
A new study has shown promising results for non-invasive blood glucose monitoring using wearable technology.

Published in Nature, the study involved 50 participants with type 2 diabetes. Researchers used a Raman spectroscopy device to monitor blood glucose levels over two days without requiring blood samples.
The device used near-infrared light at an 830nm wavelength, positioned over the participant’s thumb. It analysed the reflected light, collecting data over 50 seconds for each hand placement. The results were then compared with traditional blood sample testing.
While spectroscopy-based glucose monitoring is not new, this study introduced a pre-trained calibration model. This model, derived from a previous study with 160 participants, reduced the initial calibration stage to 10 measurements over four hours. The study found that this setup allowed accurate readings for at least 15 days.
“These results highlight the ability to reliably track blood glucose levels in people with type 2 diabetes,” the study concluded.
Although the study did not use consumer-grade electronics, the Raman spectroscopy technique is similar to what Apple has reportedly been exploring for its Apple Watch. Apple has been working on non-invasive blood glucose monitoring since at least 2017, but challenges remain in achieving reliable results and miniaturising the technology for a wearable device.
Currently, the Dexcom G7, a patch-based glucose monitor that connects to the Apple Watch, is the closest alternative for consumers.
Samsung is also developing a non-invasive blood glucose monitoring solution. In early 2025, Samsung Senior Vice President Dr. Hon Pak announced the company’s efforts to integrate AI-driven glucose monitoring and nutrition coaching into its technology.
A study in Nature tested non-invasive blood glucose monitoring on 50 participants with type 2 diabetes
Researchers used Raman spectroscopy, which analyses reflected near-infrared light
A pre-trained calibration model reduced setup time and allowed accurate readings for at least 15 days
Source: FORBES