Health IT, Patient Engagement

New study by Cardiogram confirms accuracy of diabetes detection via fitness trackers

Cardiogram's study with University of California San Francisco recruited 14,011 participants who access the Cardiogram app through Apple Watch or Android smartwear. 

Cardiogram assessed app users’ heart data for a study on diabetes with UCSF.

Medtech companies have enlisted novel approaches to detecting diabetes and measuring blood glucose levels from tears to sweat, if only to show that it can be done. But a new study by digital health startup Cardiogram with the University of California San Francisco suggests yet another potential avenue to detection: heartbeat assessment through a wearable fitness tracker.

What’s important about the study, said Cardiogram Cofounder and CEO Brandon Ballinger via email, isn’t to change the way diabetes is currently diagnosed but to reduce the number of people unaware that they have the chronic condition or may be at risk for developing it. About 7.2 million or 22 percent of people with diabetes are undiagnosed, according to data from the Center for Disease Control and Prevention.

“By continuously monitoring using non-invasive wearables, we can identify when you’re at high risk of undiagnosed pre-diabetes or diabetes, offer you a free A1c test, and then guide you to clinically appropriate treatments,” Ballinger said.

The study recruited 14,011 participants who access the Cardiogram app. Although the app is available through Apple Watch or Android smartwear, the study only involved Apple Watch users.  The web-based, IRB-approved study was run in partnership with the cardiology department at UCSF. Participants completed a medical history, including previous diagnoses, blood test results, and medications. A mobile app integrated with HealthKit continuously stored, processed, and displayed participant heart rate, step count, and other activity data, according to a description of the study.

The study was part of the larger online Health eHeart Study by UCSF, which has enrolled more than 100,000 participants worldwide.

Using its deep neural network called DeepHeart, Cardiogram assessed 200 million heart rate and step count measurements to predict who had diabetes and who did not. The accuracy rate was 85 percent.

So how can wearable data be used to detect diabetes? Cardiogram Cofounder Johnson Hsieh noted that in the early stages of diabetes, the pattern of heart rate variability shifts.

“In 2015, the Framingham Heart Study showed that high resting heart rate and low heart rate variability predicts who will develop diabetes over a 12-year period. In 2005, the ARIC study showed that heart rate variability declines faster in diabetics than non-diabetics over a 9-year period.”

The study findings were presented at the AAAI Conference on Artificial Intelligence.

Earlier studies by Cardiogram have focused on three other prevalent but undiagnosed conditions associated with cardiovascular risk: high cholesterol, hypertension, and sleep apnea.

Christina Farr of CNBC, noting that the participants who were confirmed to have diabetes and pre-diabetes were already aware they had the condition, wondered on Twitter what would be the false positive/false negative rate in a randomly selected population. On Cardiogram’s to do list, it would be interesting to see an assessment of its technology on people who are uncertain if they have the condition.

Looking ahead, at least in the near future, Cardiogram’s Bollinger said the startup had no plans to seek FDA clearance for its diagnostic approach.

“Right now, we’d always use an existing FDA-cleared (or CLIA-waved) test to confirm the diagnosis. I think that, until consumer wearables have more sensors, that’s the safest approach. However, down the road, if the Apple Watch Series 5 has a glucose sensor, then I could see us trying to perform diagnosis with a consumer wearable and seeking FDA clearance for that.”

Photo: BrianAJackson, Getty Images

Shares1
Shares1