Can a watch measure glucose? What wearable PPG research actually shows

"Can my watch check my blood sugar?" is one of the most common questions about wearables, and the answer today is no. This page explains why, what the research has actually tried, and what PPG can contribute to diabetes research without pretending to be a glucose meter.

On our map of wearable PPG research, diabetes and metabolic health form a cluster of 10 studies. The map labels this area early, and this review keeps that framing. For a vendor-neutral look at consumer devices, see this guide to glucose monitor watches.

The short regulatory answer

In February 2024 the U.S. Food and Drug Administration issued a safety communication titled "Do not use smartwatches or smart rings to measure blood glucose levels" 1. It says the FDA has not authorized, cleared or approved any smartwatch or smart ring intended to measure or estimate blood glucose values on its own. It warns that inaccurate readings could lead people with diabetes to take the wrong dose of insulin or other glucose-lowering drugs, with serious consequences.

The same notice draws an important line. Smartwatch apps that display data from an FDA-authorized continuous glucose monitor, which uses a small sensor under the skin, are not the problem. The warning covers watches and rings that claim to measure glucose without piercing the skin.

Glucose measurement belongs to fingerstick meters and continuous glucose monitors. Wearable PPG is a research topic here, not a measurement option.

Why light struggles with glucose

PPG works by shining light into the skin and measuring how much returns. The pulsing part of the signal follows changes in blood volume with each heartbeat. That is great for pulse rate and useful for blood oxygen, because hemoglobin absorbs red and infrared light strongly and differently depending on whether it carries oxygen.

Glucose is different. It is a small molecule at low concentration, and its optical signature is weak and overlaps with water, fat, proteins and other tissue components. Skin thickness, temperature, hydration and blood flow all change the light path too. A review of non-invasive glucose technologies grouped approaches into optical, microwave and electrochemical methods and described accuracy, calibration and stability as major unsolved challenges 2.

Why the confusion persists

If no watch can measure glucose, why does the question keep coming up? There are a few reasons.

First, unauthorized devices are sold. The FDA notice says watches and rings claiming non-invasive glucose readings are made by dozens of companies, sold under many brand names and often found through online marketplaces 1. They can look polished and show a number on the screen, which makes them easy to trust.

Second, a real and useful product sits right next to them. Many people with diabetes see their glucose on a watch every day, because the watch displays data from a continuous glucose monitor worn on the arm or abdomen. The watch is a display, not the sensor.

Third, research headlines can run ahead of the evidence. A lab paper reporting close agreement in a small group is easy to read as a solved problem. It is a first step, and most first steps in this field have not survived larger testing 2.

What engineering studies have tried

Some groups have tried to estimate glucose directly from PPG features using machine learning. In one chip design paper, the authors built a single-wavelength near-infrared PPG system with an on-chip regression model and reported glucose estimates for 200 people that agreed closely with reference values 6. The same group showed a prototype system the year before 7.

These results are interesting as engineering. They are not validation. The models were developed and tested by the same team, often on the same population, under controlled conditions, and they have not been confirmed in large independent trials or against continuous glucose monitors over days of real life. A systematic review of 78 studies using PPG or ECG in diabetes care described glucose estimation as one of three research aims and highlighted differences in data size and method across studies 3.

For a broader overview of this work, see this summary of PPG research in diabetes.

What PPG research on diabetes actually studies

Most of the map's diabetes cluster is not about glucose at all. It asks how diabetes changes the blood vessels and the nervous system, and whether those changes show up in the pulse.

Vessel function. An early study used finger PPG to measure pulse wave reflection and gave a drug that relaxes blood vessels through the vessel lining. The response was blunted in people with type 2 diabetes, while the response to a drug that acts directly on vessel muscle was preserved 5. That pattern points to changes in endothelial function.

Arterial stiffness. In 159 men who had an oral glucose tolerance test, a second-derivative PPG index of arterial stiffness was associated with age and with glucose two hours after the challenge 8. A study of 29 people with type 2 diabetes and 29 age-matched controls found that decomposed finger pulse features differed between groups and that one feature was associated with HbA1c 9. The same team looked at fingertip pulse features in people with diabetic peripheral neuropathy 10. See the wearable blood pressure research review for more on vascular aging.

Autonomic nerves. Slow fluctuations in the PPG signal are driven by the sympathetic nervous system and are usually closely matched between the left and right hands and feet. In 35 people with diabetes and 33 without, that left-right match was lower in the diabetes group and fell further with disease duration 11.

Classifying diabetes from pulse shape. Toe PPG features from 58 healthy people and 83 people with type 2 diabetes were used to train a classifier 12. At a much larger scale, a deep learning model used smartphone camera PPG recorded from a fingertip in 53,870 people. It separated people with and without diabetes with moderate accuracy, an area under the curve of about 0.77, and performed similarly in a second cohort of 7,806 people. Adding age, sex, race and body mass index raised the area under the curve to about 0.83, and the model's score was associated with HbA1c 4. That study is the strongest evidence that the pulse carries diabetes-related information. It still describes a population-level signal, not a test for an individual, and not a glucose reading.

Blood oxygen caveat. Diabetes can also affect other PPG-based measures. In 261 people with type 2 diabetes receiving oxygen therapy, higher HbA1c levels were associated with pulse oximetry overestimating arterial oxygen saturation 13. Teams studying people with diabetes should keep this in mind when they interpret blood oxygen. See SpO2 monitoring for more.

One study from the source review, a logistic regression classification paper published in 2020, has since been retracted. We removed it from the map cluster and its counts, and we do not cite it here.

How research teams use PPG around metabolic health

PPG cannot stand in for glucose data, but it can sit next to it. Common designs include:

Pair with a glucose monitor. Use an authorized continuous glucose monitor for glucose, and use wearable PPG for heart rate, heart rate variability, sleep and activity context. This lets researchers study how glucose patterns relate to sleep, stress or exercise.

Study vascular and autonomic change over time. Pulse shape and variability measured at rest, under the same conditions, can be tracked over months in a metabolic program or trial.

Track physiology during treatment. Programs using metabolic medications often want heart rate, variability and sleep context. See monitoring physiology during GLP-1 therapy.

Keep the raw waveform. Vascular and autonomic features need beat-level data. As an example, Sensor Bio V2 records raw green, red and infrared PPG at 25, 50 or 100 Hz and derives metrics such as heart rate and heart rate variability, and partner teams can export the waveforms. V2 is a general wellness device and is not FDA cleared. It does not measure, estimate or report glucose, and it is not intended to detect or manage diabetes.

How these signals are measured: the science of PPG. For every research area, return to the wearable PPG research map.

For research teams

Put the raw signal in your study or product

Running a metabolic study alongside a glucose monitor? We can help you plan raw PPG capture and export.

Talk to Sensor Bio

Limits and open questions

• No wearable PPG device has been validated to measure glucose, and none is FDA authorized to do so 1.
• Direct glucose estimation papers are small, often single-team, and not independently confirmed.
• Most diabetes-related PPG studies used finger or toe sensors at rest, not wrist wearables in daily life.
• Classification studies describe group differences and cannot diagnose diabetes in an individual.
• Diabetes can bias other optical measures, such as blood oxygen.