Artificial Pancreas Makers Race to Market

Strolling through a nature preserve near the Old North Bridge in Concord, Mass., where “the shot heard ’round the world” began the American Revolutionary War, Ed Damiano is talking about another revolution, this one in the care of his son’s Type 1 diabetes. Call it the insulin shot heard ’round the world, delivered not with a handheld syringe, but automatically, from a computer-guided pump: the diabetic answer to the driverless car.

“In the fall of 2017, David heads off to college,” says Damiano, a professor of biomedical engineering at Boston University. “I want to have a version of our device approved by then.”

Damiano’s device is a technological tour de force that diabetics and their families have been clamoring for, and researchers have been studying, for decades. With hardware as complex as any medical instrument and software as convoluted as a stock-trading algorithm, the device replicates with aluminum, glass, silicon and plastic what a functioning pancreas does with biological tissue: continuous, autonomous, near-perfect control of blood-sugar levels.

His bionic pancreas — and similar artificial devices in the works — is as small as a cell phone and worn externally, with spaghetti-thin tubing inserted just under the skin. It releases insulin without any action required by the user, in the right amount, at the right time.

That’s a big deal for the estimated 1.5 million people in the United States with Type 1 diabetes because the amount of insulin they need at any given moment is always changing — day to day, hour to hour, even minute to minute. Insulin is the hormone that allows your body’s cells to absorb glucose, the gasoline that makes cells go. No insulin, no life.

The problem with replacing the lost insulin is that, unlike almost any other drug, a fixed daily dose is nearly impossible to set. The need rises and falls based on how much glucose is in the blood, which, in turn, is determined by how much carbohydrate has been eaten and how much exercise taken. Too little insulin, and blood-glucose levels spike high, which over the long term can lead to heart disease, kidney failure, blindness and amputation. Too much insulin, and blood-glucose levels can fall so low that brain cells lack the fuel they need to function. Confusion, coma and death follow unless the diabetic eats some glucose. That’s why people who take insulin have to prick their fingers for blood multiple times a day to test their glucose level and then figure out how much insulin to take. Damiano intends to make most of that hassle, and most of the danger, disappear. It’s a technical fix that, while not a cure, would be transformative nonetheless.

But don’t mistake him for a lone ranger. Damiano has lots of competition from other tech-geek dads of diabetic children. Some of them have already formed companies to compete with the Goliath of the diabetes technology field, Medtronic, which is developing its own device. Most of them think Damiano’s approach is too risky and complicated because his device will pump not only insulin but also the glucose-raising hormone glucagon. But then, Damiano has his own doubts about their approaches, too.

It’s kind of heartbreaking to hear these extraordinary innovators pick apart each other’s work as they race to finish their various devices, complete clinical trials big enough to gain FDA approval and ramp up production with sufficient support personnel to help diabetics actually strap one on by 2017 or 2018. But it’s also inspiring to realize that the story of the artificial pancreas has finally turned a scientific corner. It is no longer about a single researcher living on a prayer of finding a breakthrough treatment. The dreaming and the research are giving way to implementation. And at this point, all that remains to be seen is which researcher, and which technology, will come to market first.

Long considered the jet pack of diabetes care, the artificial pancreas, first described in a medical journal in 1972, has been promised so often that many diabetics had given up on the possibility that it would ever arrive in their lifetime. That included me: I was diagnosed with Type 1 at age 18, two days after Thanksgiving in 1975.

A few pioneering researchers, Damiano among them, finally began testing prototype artificial pancreases in humans in 2008. But progress toward an FDA-approved product remained in an academic, regulatory and commercial quagmire until an upstart technology millionaire named Jeffrey Brewer came along. Brewer, whose son was diagnosed with Type 1 diabetes at age 7, was so frustrated with the turgid pace of development that in 2004 he offered $1 million to JDRF (formerly known as the Juvenile Diabetes Research Foundation) if the group could bring together researchers, manufacturers and the FDA to talk seriously about getting an artificial pancreas to market. Not only did JDRF take him up on his offer, it eventually named him president and CEO of the foundation. In 2010, Brewer predicted that we’d see an artificial pancreas approved and on the market within five years.

Well, not quite.

By late 2014, despite dozens of randomized trials showing that they worked brilliantly, no approved device was even on the horizon. So frustrating was the lack of progress that parents and a few adults with diabetes started hacking their own devices. Some figured out how to hack their continuous glucose monitoring (CGM) systems to send glucose readings to their cell phones.

Then rumors flew that some dad had put his son on a homemade artificial pancreas — but he didn’t want to tell anyone until he had a way to distribute it ethically. Nobody knew if the story was true; I described him in a magazine article as being like Bigfoot: widely rumored, never seen.

In this case, however, Bigfoot turned out to be real.

A Manhattan stock-trading programmer named Bryan Mazlish, who had made millions on computer algorithms he’d designed to autonomously buy and sell stocks, confessed that he was the guy. In November 2014, Mazlish teamed up with Brewer to form a company to commercialize a much-improved version of the artificial pancreas he had built. To do so, Mazlish shut down his stock-trading business, Brewer resigned from the leadership of JDRF, each of them ponied up $1 million, and Brewer threw himself into raising millions more from venture capitalists. They persuaded Medtronic’s chief engineer for diabetes, Lane Desborough, to quit his job and join them. And the formal name they gave their company — I couldn’t help but smile — was Bigfoot Biomedical.

“Think about how many smart mathematicians and banks would like to create their own algorithms to make money automatically off the stock market like Bryan did,” says Aaron Kowalski, chief mission officer and vice president of research at JDRF. He was trying to explain why he believes that Mazlish and Brewer will beat out all the academics who have devoted their careers to studying the artificial pancreas.

“I am not belittling the abilities of the academic investigators,” Kowalski says. “JDRF has funded them. They have brought us to where we are today. But they don’t understand how to commercialize something. Bigfoot is very different. Jeffrey ran two successful startups. Bryan’s algorithms had to have impeccable reliability so he didn’t lose millions of dollars on Wall Street. Before he joined Medtronic, Lane Desborough used to run chemical processing plants with control algorithms where, if they screwed up, workers would die. These guys know what they’re doing.” Ironically, the first person to try Mazlish’s homemade artificial pancreas — his wife, Sarah Kimball, who also has Type 1 — didn’t initially see the point of the contraption.

“I’m very Type A,” confesses Kimball, a pediatrician. “I tested my blood-sugar level probably 12 times a day. I always tried to make diabetes appear effortless. I didn’t want people to think of me as sick.” Kimball was so good at managing her diabetes that Mazlish had no idea how complicated the disease truly was until their second child, Sam, was diagnosed with Type 1 at age 5. That’s when Mazlish wondered if there was a better way to care for Sam other than waking up every two to four hours to check his blood-sugar level and calculate how much insulin to pump or glucose to slip under his tongue.

In February 2013 — after months of coding, tinkering, reading academic studies and interviewing his wife to understand her intuitive decision-making — Mazlish asked if she was ready to try his homemade artificial pancreas.

“I knew once he was willing to put it on me, it would be totally safe,” Kimball says. “He is the most careful and circumspect guy there is.”

“One of the skills I brought to this,” Mazlish says, “was a strong experience in developing models for simulating stock trades, where it’s important to develop robust simulations to test out strategies and ideas before implementing them. In stock trading, you’re putting a person’s money at risk; in diabetes, you’re putting a person’s life at risk.”

That first test day, Kimball watched in wonder as the pump sprang to life, automatically pumping insulin to cover her blood-glucose spike after breakfast, then cutting off the insulin after she exercised. Most impressive to her: “I woke up with a blood-glucose level every morning between 90 and 120,” she says, quoting numbers that are well within the range of normal.

Convinced that the device was safe enough for Sam, Mazlish and Kimball found they could let their son eat and play like a regular kid, sleep through the night without constant checks and go to a camp without an on-site nurse. Not that the device was perfect: During the daytime, especially, it still required Sam and his mom to occasionally drink juice or take glucose tablets when their blood-glucose levels dipped too low for comfort. But lows happened more often without the device, and with it they were less severe because the algorithm calculated the curve of their glucose levels when it was still in a safe range and switched off the insulin pump once it saw a low in the offing.

Mazlish didn’t know it at the time, but the DIY technology that changed the lives of his diabetic wife and son would soon change his own. In late 2014, he met Brewer. They got to talking and decided to throw their careers, reputations and a significant part of their fortunes together to turn a diabetic dream into a business. The fact that academics and manufacturers had already been working on an artificial pancreas for years didn’t worry them in the least.

Sitting in his office at Boston University, Damiano is equally unworried about the competition. Fit and trim in a black T-shirt sporting a “Go Bionic” logo he designed, Damiano exudes both youthful exuberance and scientific authority — a boyish face topped by a mantle of thick, gray hair.

In an interview in the spring of 2015, Damiano confidently predicted: “We’re going to be running a pivotal trial by the end of the year,” referring to the type of clinical trial designed specifically to meet the FDA’s standards for getting a device approved. Damiano and his research partner, Steven Russell, an endocrinologist at the Diabetes Research Center at Massachusetts General Hospital, have been developing their technology for over 12 years. “We’ve tested our device in people ranging from 40 pounds to 300 pounds, children to adults, during night and day, with no restrictions on food or activity.”

By contrast, he says of Brewer and Mazlish: “They have no clinical data. They have never run a study. They have two people on a device, Bryan’s wife and son. It’s never been peer reviewed. I do believe they have talent, but the problem is that it’s a very iterative process to develop a set of algorithms that work not just in two people but in hundreds of people. I don’t think they understand the challenges.”

Damiano and Russell’s greatest success so far, one that drew glowing media coverage, was a 2014 study published in the New England Journal of Medicine. It showed that both adults and adolescents with Type 1, eating and exercising as they liked outside of a hospital setting, had better control of their blood-glucose levels and fewer incidents of severely low glucose levels (hypoglycemia) during five days on their bionic pancreas compared with their usual care.

“Our system is totally automated,” Damiano says, contending that other devices under development require all kinds of information to be entered before using them. “Literally you just type in your body weight, and you’re ready to go.” Actually, like most of the other devices, his must also be regularly calibrated by entering blood-sugar numbers obtained from a finger-prick test. But in contrast to current “dumb” pumps and ordinary injections, which require diabetics to estimate how many carbohydrates they’re eating and how much insulin they need to cover each gram of carb, his device is what he calls “diabetes without numbers.”

“You look at your meal, and you tell the device if it’s a snack, a small meal, an average meal or a large meal,” he says. “The device comes to learn what you mean by each of those.”

Although the device has a button to alert it that you’re planning to exercise, Damiano says it will work fine if you don’t press it because the system pumps both insulin to lower blood-glucose levels and glucagon to raise those levels quickly when necessary. “For it to be a truly autonomous system,” Damiano says, “it needs some way to raise blood-glucose levels other than just turning off insulin, [the effect of] which is ever so slow. Glucagon raises your level fast.”

By 2016, however — six months after our first meeting — the most notable variables to be lowered dramatically are Damiano’s projections for when his device will be tested, approved and on the market. Rather than beginning his pivotal trial by the end of last year, he now predicts it won’t start until spring 2017 and won’t be completed until that year’s end. Damiano’s dream of sending his son, David, off to college with an approved device is no longer in the cards — at least not with the device Damiano and Russell designed.

What’s rarely mentioned in media coverage of Damiano’s work is that glucagon is approved by the FDA only for emergency use, not for the multiple daily microdoses used in the bionic pancreas. What’s more, glucagon costs between $150 and $200 per vial and must be replaced daily. Damiano and Russell are working with Xeris Pharmaceuticals to get approval for a version of glucagon that can last for months, and it will be tested for FDA approval in their upcoming pivotal trial. At the very least, say other researchers, using a two-hormone system adds unnecessary cost, risk and complexity — particularly because insulin by itself, without the addition of glucagon, has already been shown to work well in other studies of artificial pancreas systems.

“We use only insulin and, it may surprise you, but we see less hypoglycemia than in the published studies of the device with glucagon,” says Boris Kovatchev, director of the University of Virginia Center for Diabetes Technology. For a dual-hormone system, like Damiano and Russell’s, to work, “you have to make the assumption that there will be a stable glucagon available and that the glucagon never fails. If the glucagon arm fails, that’s putting the person at risk,” he says. Kovatchev has racked up nearly 184,000 patient-hours, and counting, of clinical trials involving his version of an artificial pancreas — the time equivalent of one person wearing his device for more than 20 years. His system has also been studied at medical centers in the United States, France, Italy, Israel and Amsterdam.

At least two other academics are also in the race to bring an artificial pancreas to market. Roman Hovorka, director of research at the University of Cambridge Metabolic Research Laboratories, has published more than 30 clinical studies since 2010 of what he calls a closed-loop system. Moshe Phillip, director of the Institute of Endocrinology and Diabetes at Schneider Children’s Medical Center of Israel, has treated some 1,700 patients with his own system.

All three of these academic competitors — Hovorka, Kovatchev and Phillip — have already partnered with companies to commercialize their efforts. Hovorka and Phillip are with Medtronic, the current industry leader in diabetes technology; Kovatchev is with a startup named TypeZero Technologies. Insisting that he doesn’t need a traditional company to commercialize his device, Damiano founded a “public benefit” corporation in October, the kind normally used to run transit systems and utilities. The firm, Beta Bionics, quickly secured $5 million in funding from Eli Lilly and Co., the pharmaceutical giant. Damiano serves as CEO but remains a professor at Boston University, where he continues to seek research grants from NIH.

That kind of approach drives Brewer and Mazlish nuts.

“They do science. We do business,” Mazlish says. “This isn’t to deride the academics, but they’re still filling out NIH grant applications and publishing papers. We’re commercial to the core.”

Brewer states their position bluntly: “We’re all in. We succeed if we get approval by the FDA, create a business and ship products to people. If we can’t deliver product, then we fail. The academics still succeed by developing better algorithms, getting papers published. They’re not on the line to do the one thing that needs to be done: to get this technology into the hands of the people who need it.”

Mazlish and Brewer fully understand they will need to complete a pivotal trial involving hundreds of people with Type 1 to gain FDA approval for their device. And beyond the $1 million they each contributed to Bigfoot as seed capital, they had raised another $15 million from venture capitalists by February of this year and were expecting another substantial infusion within months.

Their ambitions, however, go beyond raising enough capital to build a device and get it approved. Because endocrinologists tend to be the only doctors familiar enough with diabetes technology to prescribe pumps and CGMs to people with Type 1, and because endocrinologists are in such short supply that many diabetics never see one, Bigfoot intends to set up a one-stop shop that will connect certified diabetes educators with patients to explain how the artificial pancreas works.

But in that regard, Medtronic has a head start, with some 3,000 diabetes educators already teaching patients how to use its pumps and CGMs. Mazlish and Brewer deride the company as a lumbering giant moving so slowly that their partner Desborough left to join Bigfoot.

But Francine Kaufman, chief medical officer of Medtronic’s diabetes unit, disputes that point. “I don’t get that,” she says. “We’re the only ones in a pivotal trial. We already have patients enrolled. We have a huge call center and a huge cadre of certified trainers.”

Like Damiano and Russell, Bigfoot has been pushing back the projected date for its pivotal trial. After initially saying it would begin by the end of this year, Brewer now says it won’t start until 2017. But he sticks by his original projection for the device to be approved and on the market by late 2018 or early 2019. Even so, Kaufman says, “there’s no way they’re going to be first to market.” She estimates Medtronic will have a “hybrid” artificial pancreas — hybrid because patients will still have to alert the device when they’re going to eat or exercise — by the end of 2018. “Maybe even earlier,” she adds.

For its part, the FDA welcomes anyone seeking to bring an artificial pancreas to market. “We are encouraging people to come in and to get to the place where they are ready to do their pivotal studies and to submit a premarket approval application,” says Stayce Beck, chief of the FDA’s Diabetes Diagnostic Devices Branch. “I can tell you that several groups have come forward, but I can’t confirm who has or how many.”

No one can yet be sure which group will cross the finish line first, or which device will prove most popular. What matters is they are all in the game, each one making contributions, even as they fight to turn the artificial pancreas into a 21st century gadget as life-changing as the iPhone — and as life-saving as insulin.

When it was first isolated in 1921 by Canadian researchers Fred Banting and Charles Best, and injected soon after into diabetic children, insulin was rightly celebrated as a 20th-century miracle drug, but wrongly deemed a “cure” for the disease.

Banting and Best did not foresee the extraordinary trouble caused by taking the hormone from an external source rather than having it secreted as needed by the pancreas. It’s nearly impossible to set a daily dose.

Back when I was diagnosed with Type 1 diabetes in 1975, little had changed in over half a century. Insulin had been purified compared with the mucky stuff that Banting and Best first isolated, and an additive was found to make a longer-acting version that could last up to 12 hours.

The hospital nurse who taught me how to inject myself told me with great enthusiasm that I was lucky because disposable plastic syringes had just come on the market, so I wouldn’t need to boil and sterilize a glass syringe three times a day. Yippee!

So much for “progress.” Every morning, I had to inject a set dose of long-acting insulin and a small dose of short-acting insulin, both of them derived from a pig or cow pancreas. At dinner, I was supposed to take a second injection of short-acting insulin, and at bedtime, I was to take a third injection, again of long-acting.

And because there was no way, outside of a laboratory, to check one’s blood-glucose level, I was supposed to match my food intake to the insulin schedule by eating the exact amount of food at exactly the same time every breakfast, lunch and dinner, with regular snacks between meals.

In practice it was all madness. I was an 18-year-old English major with much more interest in girls, graffiti and guitars than in becoming a dietary robot.

As a result, three times in the first five years after my diagnosis, I found myself waking up in the back of an ambulance, where medics had just given me an injection of glucagon, the hormone that prods the liver to instantly release its stored-up glycogen, a dense form of glucose.

And, lest you think me a special kind of idiot, I can assure you that similar events pockmark the lives of nearly everyone with Type 1, unless they’re really, truly obsessives who follow a fantastically strict diet-and-exercise routine (and are just plain lucky).

The first big breakthrough since the discovery of insulin came with the advent of home blood-glucose testing in the early 1980s, allowing diabetics to check their sugar level in a minute or two with a drop of blood on a reagent strip. Now I could quickly see if I was running high, in which case I injected a bit of insulin, or low, in which I’d grab a sip of juice or anything else sweet.

A decade later, insulin pumps became widespread. They released the hormone via a thin catheter injected under the skin and changed every three days. Instead of having to whip out a needle every time insulin was needed, diabetics could just press a button on the pump.

The next big step came in 2005, with the advent of the continuous glucose monitor (CGM). Like the pump, a CGM includes a slender filament injected under the skin and kept there for a period of days, but in this case it’s a minuscule electrode that senses changes in conductivity due to glucose levels. Instead of having to prick your finger for a drop of blood to place on a strip inserted into your blood-glucose meter, you could just glance at your CGM monitor to see if you’re high or low, rising or falling.

Taking appropriate corrective action with a press of your insulin pump or a handful of gummy bears couldn’t be easier, right?

Wrong. The overriding problem with all this technology is that it sucks up that most precious of resources: mindshare. Like sheep in a meadow, the pump and CGM require the good diabetic shepherd to pay ceaseless attention. What’s my current reading? How many carbs are in this sandwich? How much insulin should I pump?

Calculating these things is not easy (especially, of course, when you’re asleep, not to mention working, watching a baseball game, out on a date, going for a run or otherwise engaged in life).

And the job is made inhumanly difficult because even the fastest-acting insulin currently on the market takes over an hour to reach peak effectiveness, while a meal begins raising glucose levels within minutes. It’s as if you have to swing your baseball bat nearly an hour before the pitcher throws the ball.

But wait a minute. What was that word I mentioned earlier? Calculating. What could possibly do a better job of calculating than a human does? Hmmm. Wait a minute, it’s on the tip of my tongue . . .

[This article originally appeared in print as "Priming the Pump."]