Tandem’s Control-IQ Explained

Last fall,  I wrote about how glad my husband and I were that our daughter, Bisi, had pushed to switch to the t:slim X2 pump. And that was before we’d been able to download and start using Tandem’s Basal-IQ system, which uses data from Dexcom’s G6 CGM to shut off all insulin when the user is predicted to go low.  

Basal-IQ has been great for Bisi, but I would say that Mark and I have benefitted from it just as much. It seems to do its best work when Bisi is asleep, and there are fewer variables to contend with: she’s not eating, and she’s not exercising. Whereas we used to have to get up fairly frequently to either turn off Bisi’s basal rate and/or give her some juice, it’s now very rare that we need to wake up to address lows. On some mornings, the screen of Bisi’s pump will show that her basal rate shut off by itself perhaps a dozen times, while her blood sugar stayed 90 even as she slept. The system is helpful during the day, too, and takes the edge off some of her lows. But if she has mistimed a bolus, or given too much insulin for a meal, Basal-IQ is an improvement, but not a total fix for the lows. And of course, it doesn’t even address the highs.

That’s where Tandem’s next big update, Control-IQ, comes in.  Scheduled to be available during the second half of this year, Control-IQ is an advanced hybrid closed loop system that will both decrease insulin when blood sugar is going low, and increase it when it’s going high.  As Tandem’s Betsy Dokken, Director of Clinical Affairs, and Garret Marin, Director of Project Management, explained to me, Control-IQ is not the next generation of Basal-IQ. It’s a different system, based on a different algorithm, with a lot more capability.

Basal-IQ is a simple algebraic algorithm that uses the last four blood glucose readings from the Dexcom CGM to predict what the glucose level will be over the next 30 minutes; if the number is predicted to go below 80, the pump will shut off insulin delivery. By contrast, Control-IQ is a complicated algorithm based on calculus and differential equations that is doing “much more in the background than Basal-IQ,” explains Marin. Or, as Dokken puts it, “Control-IQ is more similar to how the human body manages blood glucose, and considers if you’ve eaten recently, if you’ve exercised recently, and other inputs to try to mimic human physiology.” 

The algorithm was developed by scientists at the University of Virginia, licensed by the diabetes technology company TypeZero, and has undergone more than 250,000 hours of clinical testing. “They had tested that algorithm pretty robustly, so we took it and embedded it in our pump with very few changes,” says Dokken. If a person downloads Control-IQ, the algorithm will be layered on top of the individual pump settings. As Marin describes it, the algorithm is designed to keep you in a range between 112.5 and 160. When it predicts that you will go below 112.5, it can either decrease or turn off the basal rate, as needed (Basal-IQ only turns the basal rate off or on). If it predicts you’ll go above 160, the system will first raise the basal rate, and then, if necessary, will give a correction bolus (this automatic bolus can only be given once an hour, and is programmed to be 40% less insulin than the pump’s settings call for).  Dokken describes the process this way:   “Every five minutes Control-IQ makes a decision. If glucose is in the target range, it delivers the insulin according to the settings. If out of range, it will give more or less insulin. It makes a different decision every five minutes, so it can deliver a different basal rate twelve times an hour.”

The system has two additional settings built into it: one for when you’re sleeping, when the target range is narrowed to between 112.5 and120 mg/dl and the algorithm doses insulin more aggressively. And the other for when you’re exercising, when the algorithm is more conservative in terms of giving insulin. 

In a limited sense, the algorithm can learn over time. If the algorithm determines that the amount of insulin you actually use is more or less than what’s specified by the settingson your pump, the algorithm will adjust. “So it learns based on total daily dose and it changes based on that value,” explains Marin.

The overall goal of Control-IQ is to make living with T1D easier—to predict lows or highs and avoid them, without the user having to do much of anything.

For people who manage their diabetes super tightly and achieve an A1c in the high 4s or low 5s, and for people who have a DIY pump and are used to being able to control every aspect of it, Control-IQ might be a bit frustrating.  “I heard from one person who had an A1c of 5.2 that their A1c became slightly higher.  So there is the potential that they would be pushed to a higher target than they would want,” says Marin.   Also, the target ranges are set, and can’t be changed by the user.

But for the majority of people with type 1, he says, Control-IQ could be a clear benefit, either in terms of the amount of time they spend managing their disease, or in terms of a lowered A1c—or maybe even both. “I think the people in the sevens and sixes will get a lot of value out of this. I think they’ll be able to keep this level with a lot less work,” he says. Marin says there are definitely aspects of the system that “could be improved for the future. And I’m not going to say it’s going to be perfect for everyone.” But he does believe it could significantly improve the lives of those who use it.

Control-IQ is now in pivotal clinical trials, with 168 people enrolled at seven sites across the country, including at UVA, the Mayo Clinic in Minnesota, and the Joslin Diabetes Center in Boston.  Results are still pending, but will likely be submitted to the FDA in the next several months.Meanwhile, Dr. Sue Brown, an associate professor of medicine at the University of Virginia, recently released results from a smaller study, Project Nightlight, which compared people using Control-IQ versus people using both a pump and CGM without any of the Control-IQ features. The people using Control-IQ did better on all measures, including time in range (73.5% versus 59.3%)and time spent below 70 mg/dL (1.5% versus 3.4%).A trialconducted in 2017,at two different ski camps for adolescents with T1D, also showed a benefit for those using the closed-loop system. The closed-loop group spent 71.3% of their time in range versus 64.7% for the control group, which translated into an extra hour and forty minutes in range per day.

And it doesn’t say so in the study results—this isn’t what the study was testing—but based on Bisi’s own experience skiing, I’ll bet the real benefit for the adolescents in the closed-loop group came from the idea that their pumps were working quietly in the background, raising and lowering their insulin levels without them having to do a thing. I imagine there were fewer times they needed to give corrections while waiting in lift lines; fewer times they had to stop for some glucose tabs and to turn down their basal rates before skiing down the hill; and fewer times they needed to take a break and wait for their glucose levels to come up. For someone with type 1, benefits like those are hard to measure; but in terms of living daily life, they’re very real.

Katie Bacon

Katie Bacon is a writer and editor based in Boston. Her daughter, Bisi, was diagnosed with type 1 diabetes in August, 2012, when she was six. Katie worked as an editor at The Atlantic, and her writing has appeared in The New York Times, The Boston Globe, and other publications. Katie and her husband, Mark, have two children—Bisi and her older brother, Jamie.

View Author Blog

Katie Bacon

Katie Bacon is a writer and editor based in Boston. Her daughter, Bisi, was diagnosed with type 1 diabetes in August, 2012, when she was six. Katie worked as an editor at The Atlantic, and her writing has appeared in The New York Times, The Boston Globe, and other publications. Katie and her husband, Mark, have two children—Bisi and her older brother, Jamie.