In 2005, Simon Carter and his wife welcomed their first daughter, Lucy, into the world. Two years later, Carter was dismayed to learn that Lucy had developed type 1 diabetes. Carter himself had been living with diabetes since his own childhood, so he knew from personal experience the challenges that lay ahead for his young child.
When Lucy started school, Carter worried about her when she was away from home. Some teachers didn’t follow Lucy’s care plans – particularly those she only saw for short stints during the day, like her swimming coach or music teacher.
Employing his background in computer programming, Carter decided to set up a web-based logging system that all of Lucy’s teachers could access. It was the first step in a nearly decade-long odyssey that would eventually lead him to launch a mobile app called PredictBGL, which is designed to mimic the predictive benefits of a closed-loop artificial pancreas.
Eventually, Carter’s innovative app would land him a spot as a semi-finalist in the 2016 Diabetes Innovation Challenge, hosted by T1D Exchange and M2D2 and supported by the American Diabetes Association and JDRF. The app would be used to improve the lives of more than 20,000 people living with diabetes across the globe.
Want more type 1 diabetes-related news stories, and the chance to help type 1 diabetes research? Take a moment to join T1D Exchange Glu now by clicking here.
Tinkering leads to a predictive algorithm
But back in the mid-2000s, Carter didn’t yet know where this tinkering would lead him. He just knew that he wanted to build a tool to keep his daughter healthy and happy as she trotted off to kindergarten each day.
He started small with a simple web-based tracking system that was a big step up from the messy, paper-based log book her teachers had been using. Modifying the website, Carter took the next logical step and added dose calculation to a system that mimicked the inputs of an insulin pump by inputting carbohydrate ratio, correction ratio, and active insulin time. With the dose calculation system on board, it was easy for Lucy to go to the nurses station every day during lunch, perform a blood test, and receive her injection.
Later, Carter started thinking bigger. What if he could use all the data on board to make predictions about future blood sugar levels? He searched academic literature and realized it wasn’t a far-fetched idea.
Carter used himself as a guinea pig, utilizing his own personal insulin profile to experiment with his dose calculations and predictions. When the prediction algorithm was finally ready, it churned out its very first prediction: based on Carter’s background data, the prediction tool guessed he was likely to experience a “low” at 11:04 am, about two hours in the future. Sure enough, at 11 a.m. Carter was at Ikea when he realized he was feeling low.
By this point, he’d developed the system to the point where it could be employed as a smartphone app, and he cautiously rolled this system out for Lucy as well. For Carter, the power of what he had built really hit home one day when he was monitoring Lucy’s profile remotely during one of her swim lessons, and realized that the app had predicted she would have a low while she was still in the pool.
He called the school secretary, who rushed over to swoop Lucy out of the pool. Sure enough, her blood test rang in at 1.7 mmol (roughly 30 mg/dL) – a dangerously low number. The app had saved the day.
The PredictBGL app hits the Australian market
Today, PredictBGL is a fully commercialized product available on iTunes and Google Play. It has been approved by Australian regulators, and Carter is currently pursuing regulatory approval in Europe. Carter’s website pitches the app as “the first system EVER to predict blood sugar levels” and as a “stress-free diabetes management” tool.
The app uses the same type of algorithm you would find in a closed-loop “artificial pancreas” system, but instead applies the algorithm to multiple daily injection therapy. For patients with controlled diabetes, the app offers blood glucose predictions up to three hours ahead of time, and those predictions are accurate 87 percent of the time.
“We are the only solution to pull everything together in a cohesive way,” says Carter. “And at such low cost, you pretty much get the same benefit.”
PredictBGL integrates with Dexcom, FreeStyle, Libre and NightScout, and offers a sharing interface so you can coordinate with your doctor, or so that parents can keep tabs on their children.
Currently, PredictBGL has about 20,000 users. Carter says a range of patients use the app – about half don’t use any devices; they use the app for dose collection. Some others use the app because they don’t trust CGMs to be accurate.
To learn more about this app, which is not yet approved for use in the United States, click here.