BBC Nov 19, 2018: Google halts glucose-sensing contact lens project
Engadget Jan 12, 2021: Startup claims its new wearable can monitor blood sugar without needles
9To5Mac Jan 25, 2021: Apple Watch blood sugar measurement coming in Series 7, claims report
These recent articles probably don’t do justice to the very long-standing interest in measuring blood sugar non-invasively. Tons of time and money has gone into trying to solve this problem, which is the tip of the spear of a bigger problem. Per the American Diabetes Association, diabetes alone affects just over 30M Americans, pre-diabetes about 90M more, and the cost of undiagnosed diabetes is over $300B. High blood glucose is highly correlated with various comorbidities that are the biggest killers in society — and one big culprit is people are unaware of their numbers. Most of us who do know the number only know from one data point per year when we get our annual physical, which in itself is a minority of Americans.
The Challenges
These are less of engineering challenges but more inherent to biology and physics:
– Glucose circulates through the bloodstream and diffuses partly towards the skin i.e., measuring your sweat is not a perfect correlation.
– Glucose is not homogenous through the blood i.e., different parts of the body do have different readings.
– Sample size matters i.e., pricking your fingers is a smaller sample than drawing a vial of blood and more prone to errors.
The Clarke Error Grid
what compounds the challenges even further is that the very gold standard of measuring through the blood has an error rate. Since 1987 in the US we have been using The Clarke Error Grid, which effectively sanctions a 20% error rate. In other words your reading could show up as 120 putting you in the category of at risk or as 80 putting you comfortably in the category of no risk, while your real number is 100. For that reason, doing repeated tests and measuring A1C (average blood glucose over 3 months) are what any doctor will recommend before taking a serious course of action.
Other Biometrics
Various labs and companies have tried circumventing the needle and collect samples from sweat, tears, urine and / or saliva. The historical challenge with all of these is when someone has very high blood glucose these methods do work but they have varying levels of success otherwise. But that hasn’t been stopping dozens of efforts from startups and corporates, ranging from Australia to Chile to South Korea. There are also many efforts on detecting glucose indirectly, say by measuring foot ulcers which is correlated with high blood sugar.
Light
Various efforts have focused on the properties of light, say diffraction (how light scatters), refraction (how light bends) and reflection (how light bounces back). One technique that has been especially favored is Raman spectroscopy, which is a non-destructive chemical analysis of how light interacts with the chemical bonds within a material to thus decide on its identity. This was at the heart of startup C8 MediSensors, which raised $60M, but couldn’t solve the variability of results from one person to another before shutting down.
Artificial Intelligence
At Tau Ventures we believe AI can help solve some of these challenges in a significant way, so much so that we dare predict effective solutions within the decade. Our belief is, as you would expect, a multidimensional approach is the most likely i.e., building models around many measurements. All of this aided by increasingly sensitive sensors, increasingly more data, and increasingly higher computational power. Case in point of where things are going, on May 14 startup OneDrop obtained CE Mark (European regulatory approval) for its glucose forecasts.
The fundamental premise of this high-level overview is that while blood sugar does have a strong genetic component, we also have meaningful agency in managing it. If you can measure it you can monitor it, if you can monitor it you can modify it.
Originally published on “Data Driven Investor,” am happy to syndicate on other platforms. I am the Managing Partner and Cofounder of Tau Ventures with 20 years in Silicon Valley across corporates, own startup, and VC funds. These are purposely short articles focused on practical insights (I call it gl;dr — good length; did read). Many of my writings are at https://www.linkedin.com/in/amgarg/detail/recent-activity/posts and I would be stoked if they get people interested enough in a topic to explore in further depth. If this article had useful insights for you comment away and/or give a like on the article and on the Tau Ventures’ LinkedIn page, with due thanks for supporting our work. All opinions expressed here are my own.
Will this technology be able to deal with the “brittle” Type I diabetic, characterized by sudden and frequent changes in blood sugar? Might it even be particularly helpful as it gathers data from multiple measures?