Wearable snake oil: The search for automatic calorie-intake tracking in fit-tech wristbands
Calories in, calories out. That’s the basic, if grossly over-simplified, equation for human metabolism. We shovel in calories every time we eat, and then burn them off during the daily grind of life—while sleeping, working at our computers, and (hopefully) getting plenty of exercise.
Activity-tracking wristbands like the Jawbone UP24 and Nike Fuel Band can do a pretty good job of telling us what’s happening on the back end of the “calorie in, calorie out” equation. They use accelerometers to monitor our exercise, then map that data against user-supplied criteria like age, height and weight. The wristbands then squeeze all this information through relatively pedestrian off-the-shelf algorithms to tell us how many calories we’ve burned.
But the “calorie in” measurement remains a wearables industry challenge. Many activity-tracking wristbands integrate food-logging tools in their mobile apps to record calorie intake, but these functions are too time-consuming and clumsy. What the fitness-tech space really needs is a breakthrough in calorie-intake tracking—some kind of sensor-laden wristband that automatically records how many calories we’re consuming in the food we eat.
“Automatic calorie intake tracking would be a holy grail,” says Daniel Matte, a Canalys analyst who focuses on the wearables industry. “There would be huge consumer demand—it’s vastly easier than manual dietary logging. But it’s probably not happening anytime soon.”
Or is it? In the last six months, two fit-tech companies claiming breakthroughs in automatic calorie-intake tracking have crawled out of the crowd-funding woodwork. Last October, Airo Health announced a wristband called AIRO that uses spectroscopic sensors to divine calorie intake through the surface of one’s skin. The company has since shuffled its executive leadership, and has refunded pre-order payments from initial crowd-funding backers. But its new CEO tells me product development continues apace.
“We wouldn’t be working on it if we weren’t confident about it,” Naman Kumar, Airo Health CEO, says.
And then there’s GoBe, a wristband announced on March 6 in an Indiegogo crowd-funding campaign. Talk about confident. Developed by a Russia-based company called Healbe, the GoBe can allegedly “automatically tell you how many calories you consume and burn throughout the day.”
But where the AIRO takes a spectroscopic approach—science that apparently piggybacks on existing (if stalled) attempts for non-invasive blood glucose measurement via simple body sensors—the GoBe wristband uses a rudimentary impedance sensor to measure fluid levels in body tissue. This fluid-level data is then married to an “advanced algorithm” to “calculate caloric intake through your skin, by reading the amount of glucose in your cells.”
It’s an incredible claim. And experts tell me it's all hokum.
“The physical reality is, this is just ridiculous,” says Ries Robinson, who has a medical degree from the University of New Mexico, a Masters in mechanical engineering from Stanford, and more than 20 years experience in developing systems for the optical measurement of body tissue. “It doesn’t work at a medical level. It doesn’t work at a practical level,” Robinson says of GoBe’s claims.
Welcome to the wearables gold rush. Time, brain power and investment capital is pouring into product development of activity-tracking wristbands, smartglasses, and even smart jewelry. Intel is rumored to be acquiring Basis, one of the biggest names in fitness bands, and Apple iWatch rumors never seem to end. But there’s also an unseemly underbelly in the wearables space: Start-ups with dubious if not outrageous product claims are avoiding traditional venture-capital sources—which probably wouldn’t give them the time of day—and are heading straight to crowd-funding to jumpstart development.
Healbe, for one, has already raised more than $766,000, far surpassing its original $100,000 Indiegogo goal. Its GoBe wristband looks like a complete, finished product—glossy and polished, with a cleverly designed companion app. The entire package appears right at home next to the Jawbone UPs and Nike FuelBands of the world. But does it even deserve to be considered a viable (let alone competitive) product?
Meet GoBe, the fit-tech miracle worker
The world knows very little about Healbe, the Russia-based company that makes the GoBe wristband. Its team is led by Artem Shipitsin, “a major developer of market solutions and new products for global brands,” according to the GoBe Indiegogo campaign page. There are apparently no medical doctors or sensor experts on the Healbe team; the closest may be Eugene Sokolov, the company’s chief scientist who “applies his background as a rocket scientist to global innovation.”
But the Healbe bona fides are actually immaterial, as the company makes bold product claims that can be reviewed on face value. According to Healbe’s Indiegogo literature, the GoBe wristband uses patented “FLOW Technology” to automatically track calorie intake. To quote the Indiegogo page: “Know exactly how many calories you’re consuming by using GoBe to read the glucose in your cells—through your skin, with no manual logging, no estimates, and no error-prone guesswork.”
The GoBe band apparently uses three sensors to feed data to the FLOW Technology algorithm. First, an accelerometer measures body movement. This is a tenable, unremarkable claim. Second, a “pressure sensor” reportedly “measures blood flow and heart rate.” There’s no scientific data that says a simple pressure sensor applied to the underside of your wrist can generate accurate heart rate data—especially in a $299 wrist-worn wearable. But let’s allow this claim to slide, as Shipitsin confirmed to me that the GoBe’s third sensor, an impedance sensor, is what generates the calorie intake data, and this is what concerns us today.
“For determine glucose, we using impedance sensor,” Shipitsin told me in an email. “We measuring impedance part of body tissue below the contact several time in a minute at the several frequencies. By measuring impedance, we don’t measure glucose directly, we measuring liquids in and out of cells. When glucose come to blood starting process of substitution water in cells to glucose (with insulin like a key). Constantly measuring impedance give as understanding of dynamics of liquids and we can calculate dynamics of glucose.”
I’ve quoted Shipitsin’s email directly, without any editing for grammar, because his claims are material to the validation of Healbe’s science, and I don’t want to miscast anything he’s shared for the public record. But, Moscow, we have a problem: Ries Robinson, who’s in the small community of experts who study the non-invasive measurement of blood glucose, says it’s scientifically untenable to deduce intracellular glucose concentration from an impedance sensor measuring fluid levels in tissue.
“Impedance has been used historically to look at the overall water content in the body,” Robinson says. “But that’s a very difficult measurement, and has largely been unsuccessful commercially. You might be able to get the glucose volume, but you’d be lacking any information on glucose concentration... The fact that there’s no peer review publication on the ability to measure glucose in the body with impedance is an enormous red flag.”
Moreover, Robinson says, intracellular glucose alone could never be a useful indicator of how many calories we’ve consumed. First, he says, we eat too many foods that are rich in fats and proteins—steak, fried fish, ice cream and guacamole, to name just a few. All these tasty treats are high in calories, but wouldn’t set off the glucose-based calorie measurement that Healbe says is the cornerstone of its product.
Second, even if we ate an all-glucose diet—just sugars and carbs—much of the glucose that we stuffed into our mouths would be absorbed throughout the body before hitting the cells underneath our wrists. This fact alone impugns glucose monitoring as a defensible tool for calorie intake tracking.
“To say that I’m going to measure glucose in the cell, and it’s going to allow me to magically understand food intake, it means all these other variables are perfectly constant,” Robinson says. “That’s not how the body works... As soon the glucose comes in, it’s either stored as glycogen or mitochondria. This is an automated reaction. It’s all in equilibrium—glucose comes in, and it gets broken down right away. The intracellular glucose concentration is largely constant, and incredibly low. There’s really no good way to measure intracellular glucose concentration.”
In summation, Robinson says, GoBe is “completely fictitious based on both medical principles as well as the physics of measuring glucose.”
Robinson isn’t the only one myth-busting Healbe’s claims. I spoke with Casper de Clerq, a partner with Norwest Venture Partners, a venture capital firm that looks for investment opportunity in the healthcare and wearables spaces. (It currently has a stake in Basis and Misfit.) He has years of experience studying what’s feasible and infeasible in the consumer fit-tech space, and he says he doesn’t believe either GoBe or AIRO are accurate enough to be useful.
“Even if blood glucose sensing was somewhat accurate by the method [Healbe] describes, this doesn’t correlate at all with food intake,” de Clerq says. “With any meal, the pancreas produces insulin, which in turn signals cells to rapidly absorb blood glucose. If indeed you are diabetic and unable to produce insulin, you will see a spike in blood glucose. However, for a healthy individual, measuring blood glucose wouldn’t reflect the amount of calories consumed because most of the glucose gets stored as fat.”
Between the damaging comments of Robinson, de Clerq, and Michelle McDonald, a clinical dietician interviewed for a Pando Daily article on GoBe, Healbe’s claims appear to be the modern fit-tech version of snake oil. Yet Shipitsin stood by his science when I raised red flags.
For example: I asked him how GoBe accounts for the fact that so many calories we consume are locked up in fats and proteins. “Protein is more like ‘construction material’ for cells,” Shipitsin replied in an email, “and without diets, contribution of protein depend on climate zone and average for U.S. is near 15 to 20 percent... If user using special nutrition like ‘low-carb’ diets, he or she must set up accordingly flags in app, GoBe change the algorithm accordingly.”
Wait, what? What’s all this talk about construction materials and climate zones? Let’s concede there may be a language barrier problem in Shipitsin’s email, and continue on.
I also challenged Shipitsin to respond to Robinson’s observation that the relationship between glucose intake and cellular glucose isn’t proportional. To this Shipitsin replied, “Yes, of course, your doctor’s absolutely right. You see the main idea—we have exactly ‘physiological technology.’ GoBe is following the process inside the body and we absolutely aligned with doctors knowledge. We preparing content to Indiegogo’s page with description of technology and internal tests of accuracy. I guess you understand, not only consumers asking us about technology.”
It would be overkill to litigate GoBe’s veracity any further here—because, after all, “internal tests of accuracy” are imminent, and it appears Healbe understands the public’s desire for serious scientific validation. (Consumers are already demanding answers on Healbe’s Facebook page and a Reddit thread.) Only time will tell if Shipitsin and his team can pull off a medical miracle (some would call it the cold fusion of wearable tech), or if Healbe will have to go the Airo Health route, and return to the drawing board.
AIRO, the original calorie-intake tracking hope
Within the context of automatic calorie-intake tracking, Airo Health, a small outfit based in Waterloo, Ontario, is the pioneer. That’s not necessarily a laudable distinction, but at least the company is contrite, and announced a refund of crowd-funded money after it determined its AIRO wristband lacks the accuracy to deliver consistent results.
It’s difficult to get a bead on exactly how the AIRO wristband works, and this makes cross-checking its scientific validity all the more difficult. I spoke with the company’s original CEO and co-founder, Abhilash Jayakumar, but he wouldn’t shed any light on the technology beyond pointing me to public statements he made when he was a member of the Airo team. But Jayakumar did explain why he left Airo Health: “The high-level reason is that I had a fundamentally different vision for the company than my co-founders, who were united on theirs,” he said.
I also interviewed Airo Health’s current CEO, Naman Kumar, another one of the company’s original co-founders. Try as I might, I couldn’t get Kumar to explain in detail how the AIRO wristband tracks calories with a spectroscopic sensor, but he did say the sensor is more of a “nutrition sensor” than a “calorie sensor.”
Based on Kumar’s brief overview and already published information about AIRO, it seems the wristband uses a miniaturized spectroscopic sensor similar to the technology once under development by C8 MediSensors, a now-defunct company that was working on non-invasive blood glucose monitoring for diabetics (the company’s Director of Optics and System Engineering, Ueyn Block, now works for Apple).
In a nutshell, Airo’s system is similar to a teeny-tiny flashlight and camera: An LED array shines different wavelengths of light through the skin, while a highly sensitive photo detector determines which wavelengths have been absorbed, and which have been reflected. The system detects the optical footprints of what Airo Health describes as metabolites, and from these, the AIRO algorithm can estimate calorie intake.
Kumar wouldn’t call AIRO a system that primarily senses blood glucose levels. He would only tell me that “glucose-sensing is one way to derive calories. It’s one of the ideas that has emerged from theoretical speculations, but it doesn’t need to be that.” He did say, however, that the AIRO system has greater success in tracking carbohydrates than other types of food: “Mostly stuff that is rich in carbohydrates—rice, bread, that kind of stuff. But proteins and fats need more time and optimization,” he said.
As for what compelled Airo Health to suspend its crowd-funded pre-order campaign, Kumar says his team hit two road blocks: It couldn’t produce reliable, scalable results beyond the lab environment, and it still has work ahead in making sure the wristband sensors always make reliable contact with the skin.
“We hadn’t done enough testing to be able to say that, ‘Yeah, not only we in our lab can use it, but you as an outside party, a third-party consumer can use it too,’” Kumar said. “We need to make sure that it works on different kinds of people. You and I are different. There are differences in how your body processes food, in how your body changes every time you eat something.”
Kumar told me that AIRO’s “expected entry to market” is mid-2015. But Ries Robinson—whose current company, Medici Technologies, develops predictive algorithms for the fit-tech industry and other sectors—said that AIRO, as it’s been described by company founders, is beyond the scope of feasibility, just like GoBe.
Robinson’s takeaway is that AIRO is focused on blood glucose, just like the GoBe system, and this is already an invalid calorie-tracking model, as it doesn’t account for dietary proteins and fats. But even more damaging, if we assume AIRO is “just” a blood-glucose sensor, it still wouldn’t have enough science behind it to produce reliable results, he says.
“Is it plausible by using one or two light emitting diodes—which looks to be the case with AIRO—could you measure glucose in the tissue with that type of optical platform? The answer is no,” says Robinson.
“We probably raised and spent $70 million trying to measure glucose non-invasively in the tissue,” Robinson said of his own experience in the field. “We had two Ph.Ds in tissue optics, and a group of individuals who were extremely good at quantitating spectroscopy, and I think we came close to realizing a device, but the hardware to do that is very sophisticated. It still has a price point even today of $5,000. There’s no way you’re doing it with two LEDs and a wristwatch.”
The case for doing nothing
Consumer tolerance for inaccuracy is already pretty high in the activity-tracking wearable space. The estimated 50 percent of users who actually stick with their wristbands seem comfortable with the fact that basic accelerometer-based reports are inconsistent from platform to platform—that if you wear three different activity-trackers on your arm, each will report different step counts and calorie-burn numbers.
What’s more important to consumers, it seems, is that calorie burn reports prove accurate within the context of their own closed platforms. And to this point, the wristband manufacturers typically publish some type of claim regarding the accuracy of their step counts and calorie-burn algorithms. Here’s one from Fitbit. And here’s one from Basis Science.
But calorie-intake tracking, based on everything we’re learning from experts, can’t even approach a minimum level of accountability. So is it even worth exploring, especially with all the negative scrutiny that’s been focused on logbook methods like MyFitnessPal and quasi-scientific methods like those mentioned in the article? Basis Science, for one, has decided to avoid calorie-intake tracking entirely—at least until the proverbial holy grail of fit-tech emerges for real.
“Today, the only ways to do track calorie intake, are, unfortunately, manual,” Basis CEO Jef Holove told me during a product demo in October 2013. “We know the user engagement just isn’t there. Consumers will say they want to do it, but yet our real-life habits just aren’t there. When we feel we’ve discovered some way of doing calorie intake that makes us feel better about it, then perhaps that gets elevated.”
In other words, Basis would be among the many companies interested in the wearable fit-tech version of cold fusion. But it also knows that today—with available technology, in a universe that must conform to the laws of physics and can’t indulge in science-fiction dreams—automatic, accurate, sensor-based calorie-intake tracking is beyond the scope of feasibility.