“It was a waking nightmare,” says John Costik of the days and weeks after his son, Evan, then 4 years old, was diagnosed with type 1 diabetes in 2012. “A 4-year-old child, who can’t [manage] any of this for himself, has an incurable, life-threatening disease,” says Costik, a software engineer in Livonia, New York. “I had all these hopes and dreams for him. I didn’t want his life changing.”
But Evan’s life would change. People with type 1 diabetes don’t produce insulin, the hormone that breaks down sugar. Based on the devices available in 2012, Evan’s parents knew that once he was old enough, he’d have to monitor his blood sugar constantly for the rest of his life. To regulate it, he’d administer insulin several times a day through a pump or a self-injection. Self-administration of insulin is a risky proposition. Give yourself too little and health complications arise over time; give yourself too much and you could die.
Until Evan could handle this on his own, his parents would manage it. Costik and his wife, Laura, quickly learned to check their preschooler’s blood sugar by pricking his tiny fingertip. They had to count his carbs and dose his insulin accordingly. But they worried most about what his blood sugar would do while he was sleeping and whether he would make it through the night.
Within a few months of Evan’s diagnosis, however, Dexcom’s G4 continuous glucose monitor (CGM) hit the market. It was revolutionary. Without the need for a finger-stick, a tiny sensor inserted under the skin painlessly sends blood-sugar readings to a small, separate receiver in its range every five minutes. The receiver connects to a computer via USB so users can view their data using Dexcom’s software.
Costik put the sensor on his son on a Friday afternoon. That weekend, he and his wife reveled in their new, real-time access to data on what was happening inside their son’s body. But when they had to send him — and his data — back to day care on Monday morning, “I panicked,” says Costik. “All that old anxiety came back. I wanted to be able to see that data while he was gone.”
Who hasn’t wished that their medication or medical device would do something that it doesn’t? For most people, it’s just an idle wish. But not for Costik. “It occurred to me, once I had settled down emotionally, that the tools we had weren’t great. We wanted to do better than that.” Costik and his wife, a manufacturing engineer, set about creating the diabetes management tools they knew were possible.
Today the Costiks are part of an international community of patients and care-givers who have hacked, crowd-sourced, and data-shared their way to the medical technology they want rather than wait for the Food and Drug Administration (FDA) to approve a new device. Some of these makers pursue FDA approval for their innovations, while others bypass it, finding other ways to get their solutions out to the masses.
While smartphones set the bar for people’s expectations of personal technology, medical devices can fall disappointingly short of what users know is possible. Why couldn’t Costik check his son’s blood sugar on his phone or online while Evan was at day care?
“Here was this device that had a USB port on it. It clearly knew how to communicate,” says Costik. “So, I started digging around Dexcom’s software application to see what I could learn.”
By design, medical devices aim to keep users from figuring out how they work, says Jose Gomez-Marquez, director of MIT’s Little Devices Lab. “Unlike any other industry, [device manufacturers] have modeled their marketing around fear. ‘If you modify your device, you might hurt yourself.’ ”
But Costik wasn’t afraid. Once he figured out how the receiver communicated with a computer, he routed the code to an Android phone hard-wired to Evan’s receiver. This “opened the loop” — it exported the data from the receiver to the phone. Through a bare-bones Android app Costik made, the phone sent the data and predictive alerts about high- and low-blood sugar levels to a cloud-based system that Costik could access anywhere.
Costik and his wife had a newfound peace of mind when they left their son at day care. “It was a huge relief to be able to see Evan’s data from work,” he recalls. “A decent portion of my motivation [to do this] was to fight off the immense anxiety I felt after Evan’s diagnosis. This allowed me to feel closer to the normal I used to know.” They could sleep at night knowing that the predictive alerts would wake them if necessary. Soon Costik was tweeting photos of himself as he viewed Evan’s blood sugar on his Pebble smartwatch while Evan was at school. In another photo he monitors Evan’s numbers on an iPad tossed in a shopping cart at Wegmans supermarket while Evan is at home, 15 miles away.
Emails and Twitter direct messages started rolling in. They were mostly from parents of other kids with type 1 diabetes who wanted the code. “I had to do one of two things,” Costik recalls. “Either commercialize it or, if I wanted to get it out there quickly, make it open source and just give it away.”
But Costik had concerns about both routes.
Dexcom had announced its forthcoming Share system that would do what Costik’s hack did. He couldn’t beat the tech giant to the market. And even just to compete, he’d have to start a company, acquire a software patent, and get FDA approval. But at that time he didn’t want to quit his day job as a software engineer at Wegmans Food Markets.
But he worried that if he gave the code away so others could hack their devices, he might be liable and the FDA might take action against him — software that reads data from a medical device falls under the agency’s jurisdiction. Still, Costik was certain that other families with type 1 diabetes needed this code.
“It had changed our lives so much,” he says. “Getting it out to other people became far more important than any financial gains.”
So, Costik and a few developers who were also affected by type 1 diabetes collaborated to launch a site — complete with a disclaimer to protect them from legal trouble — on which they’d share with the world the code to monitor blood sugar remotely. The program was called Nightscout.
Power to the People
Costik expected a couple of hundred adopters. Over 12,000 versions of the code have been downloaded on GitHub, an online code repository. Some 45,000 people combined have joined the 31 corresponding Facebook groups, including CGM in the Cloud (U.S.), Nightscout Brasil, Nightscout Israel, and Nightscout South Africa.
“The CGM in the Cloud Facebook group serves as a geek squad for DIY technology solutions,” says Joyce Lee, a diabetes specialist, designer, and a professor of pediatrics at the University of Michigan School of Public Health. “The U.S. community is answering questions about technology on the group when it’s daytime here, and the Australian community is answering questions in the middle of the night.”
Had anyone tried to sell the product, it would have taken several years and thousands of dollars to clear the FDA and hit pharmacy shelves. A medical solution by the people and for the people, “the DIY movement has democratized this technology,” says Lee.
DIYers living with various health conditions can likely identify with that democracy under the hashtag, #WeAreNotWaiting, coined by Nightscout co-founder Lane Desborough at a 2013 diabetes hackers meeting. That hashtag has since become a movement.
“That’s the power of people looking through the black box, seeing beyond it, and taking matters into their own hands,” says MIT’s Gomez-Marquez.
Diabetes Tech for the Tech-Savvy
Open-sourcing is nothing if not democratic, but is it, as Costik intended, the quickest route to getting technology to the masses?
Dana Lewis was one of the first to reach out in response to Costik’s 2013 tweets. Then a Seattle-based data analyst, Lewis, who has type 1 diabetes, couldn’t control the volume on her CGM’s built-in alarm. This meant she slept through low-sugar alerts — a life-threatening design flaw. If she could export the data to her phone, its loud ringer would certainly wake her.
Costik’s code solved Lewis’s problem. But she and her husband, Scott Leibrand, a computer network and systems engineer, didn’t stop there. Lewis, like many people with type 1 diabetes, uses an insulin pump. The pump delivers insulin as needed through a tiny tube that stays in place at all times. This eliminates repeated insulin injections, but the user still controls the pump. In fact, the user does all the thinking. People with type 1 diabetes calculate and administer insulin based on blood sugar readings or on how many carbs they plan to eat at a meal. They check and correct their blood sugar all day.
Soon after Lewis began exporting data to her phone, she and Leibrand created algorithms that would predict highs and lows and guide her in insulin dosing. Still, the last link in the chain — administering the insulin — had to be done by Lewis. But if her CGM could talk to her phone, Lewis thought, why shouldn’t it be able to send commands to her insulin pump and take her out of the loop altogether — like a real pancreas?
Through social media, the couple found Ben West, another hacker living with type 1 diabetes, in San Francisco. West had developed the code to make the devices talk to each other. Using this code, and with West’s help, Lewis and Leibrand created a rig that would automate her diabetes management. Her glucose monitor sends readings to a pocket-sized computer, such as a Raspberry Pi or Intel Edison (now discontinued), which uses algorithms the couple created to determine how much insulin Lewis needs and sends the command to her insulin pump through a tiny circuit board.
Since she started using the closed-loop system in 2014, Lewis has had far fewer low blood sugars that require her to suck down a juice box. But more than that, Lewis says, the system has given her “the peace of mind to go to sleep and know that I’m going to wake up.”
Like Costik, Lewis wanted to give her idea away. “We felt that going open source would help fill the gap between [what was available] then and a good option coming to the commercial market.”
Today, more than 500 people with type 1 diabetes have built their own Open Artificial Pancreas Systems, or OpenAPS, using instructions and code that Lewis and the community make freely available on OpenAPS.org and GitHub.
Nightscout and OpenAPS are free and open to all — all who feel comfortable modifying their medical devices, that is. To avoid legal trouble, both websites make it clear that no one will make or service your device for you. If it stops working, you have to troubleshoot and fix it yourself.
That’s why DIY systems aren’t for everyone. In a 2017 commentary in Diabetes Technology & Therapeutics, David Kerr and his co-authors noted that most Nightscout users are white, educated, and insured and would need to be social media users to access much of the information and technical support. “The worry,” says Kerr, “is that people who don’t get involved in the DIY movement will get left behind.”
The way to ensure that no one is left behind, some say, is to smooth the path of fast-paced, patient-driven DIY innovation through the FDA. After all, there are safety concerns around medical devices whose modifications aren’t FDA-approved. For starters, the FDA is responsible for evaluating the safety and efficacy of a device before it goes to market; the agency continues to monitor devices after they’re on the market. Companies must report consumer complaints to the FDA, which holds the company responsible for addressing any such issues. Adopters of DIY technology don’t have such assurances. Of course, FDA approval doesn’t guarantee a device is risk-free. Recalls of FDA-approved devices happen from time to time. In 2017 the FDA listed 32 device recalls on its website.
FDA approval is also meant to ensure the cybersecurity of smart-connected medical devices. “A scenario in which someone can hack your insulin pump to give you an overdose of insulin is a risk for anyone who has created a DIY artificial pancreas system,” says Kerr, who is the director of research and innovation at the Sansum Diabetes Research Institute in Santa Barbara, California.
But those who are not waiting have come to expect a speed of innovation that the FDA cannot deliver. Their growing movement has, however, prompted change at the agency. The FDA has begun to work with device manufacturers and researchers to ensure they design their trials correctly to meet all the necessary requirements before they start the approval process.
“We’ve started to view ourselves as a partner to work with them to get the devices to patients as quickly as possible,” says Stayce Beck, the chief of the Diabetes Diagnostic Devices Branch in the FDA’s Center for Devices and Radiological Health. This change, in part, helped get the Medtronic 670G to market three years ahead of schedule. It’s the closest FDA-approved device to a closed-loop artificial pancreas system available.
To engage DIYers, who might not have the backing of big medical tech corporations, the FDA’s Diabetes Diagnostic Devices Branch has invited makers to contact them through staffers’ personal email addresses.
“That’s helpful for [people] who feel overwhelmed by the thought of interacting with the FDA,” says Lewis. “That willingness to engage on an individual level is hugely beneficial to people who aren’t sitting in traditional companies [that have] existing formal pathways and people experienced in interacting with the FDA.”
Such changes at the agency might benefit some of the DIY-inspired artificial pancreas systems currently in the pipeline, including the legendary Bigfoot. While Lewis was crowd-sourcing code to build her artificial pancreas, she heard that there was someone out there who had already done it. “But he was very, very secretive,” she recalls. He was so secretive, in fact, that Wired dubbed him Bigfoot — many had heard he existed, but few had seen him. People speculated that Bigfoot was closely guarding secrets that stood to earn him a lot of money.
In fact, like Costik and Lewis, Bryan Maz-lish (Bigfoot), a quantitative trader living in New York City at the time, wasn’t looking to earn money from his bionic pancreas. He created a closed-loop system, which is akin to Lewis’s OpenAPS, for his wife and son. Both have type 1 diabetes. Like the other DIYers, Mazlish wanted to get the technology out to as many people as possible. But he didn’t believe open-source was the route. “Open-source will make the lives of some very engaged individuals better, like OpenAPS has, but the only way to make it broadly accessible was to go the commercial route,” he says.
Excited about his invention, Mazlish presented it to FDA officials. They were excited, too, he says, but they also took the wind out of his sails. They told him it was a class III medical device and that, as such, even if he wanted to give it away for free on the internet, he’d still have to get FDA approval.
“It’s the highest, riskiest type of device out there, and it requires many clinical trials,” he says.
For frame of reference, other class III medical devices include implantable pacemakers and replacement heart valves.
Not in a position to pursue costly FDA approval, Mazlish tried to give the idea to existing device manufacturers. But with no takers, he felt he had no choice but to keep the device under wraps until he found a way to disseminate it legally.
“I can’t tell you how hard it is to have such a life-transforming system and have to tell someone else who could benefit from it, ‘We can’t give it to you.’ ” After all, if the device were to malfunction and a child were to die, who would the parents blame?
Ultimately, Mazlish, Jeffrey Brewer, and Nightscout co-founder Desborough launched the San Francisco start-up Bigfoot Biomedical. Their closed-loop system, which began clinical trials in 2016 and will enter a pivotal trial in 2018, is expected to hit the market in 2020.
It’s the system’s DIY roots that got it down the pike so fast. Mazlish’s wife, the first to use the prototype that inspired the Bigfoot system, is a pediatrician who has had type 1 diabetes since she was 12 years old. For years, Sarah was Mazlish’s study cohort of one. He’d ask her what functions she’d like a device to have, and he’d set to work developing them and then test the result on her.
“Because we did this outside traditional medical-device development, we were doing new versions once or twice a week,” says Mazlish. “She would give me feedback. I would implement changes. The speed of innovation was incredible.” They believe that their approach will bring Bigfoot Biomedical to clinical trials ahead of the game. “We’re not using these trials to fine-tune our algorithms.”
Speed of innovation is why the DIY community, despite new commercial options like Dexcom Share and the Medtronic 670G, shows no signs of slowing. “The FDA route offers more opportunity for widespread dissemination of a product,” says Lee, “but there’s an incredible opportunity for personalization on the [DIY] side.”
That’s why Costik and Lewis haven’t traded in their hacks for out-of-the-box solutions. Both say that their systems do more than commercial options largely because of the customizing they’ve done over the years.
“I think [the DIY community] will always be able to push the envelope a little bit faster than the commercial systems,” says Lewis.
While he still uses Nightscout for his son, Costik has taken his DIY spirit to a new job at the tech start-up Beta Bionics, where he is director of mobile systems development. The company expects its bionic pancreas, iLet, to come to market alongside Bigfoot’s in 2020.
The two products will bring people with type 1 diabetes options they’ve never had before. “The more choices you can give a patient, the better,” says Courtney Lias, the director of the Division of Chemistry and Toxicology Devices in the FDA’s Center for Devices and Radiological Health. “So, if we have pathways that are set up to give people multiple devices to choose from, that’s a better scenario.”
Choices are what DIYers have been waiting — but not waiting — for. “Our community will continue to grow,” says Lewis, “until the day that there are enough commercial systems to meet everyone’s needs.”