Loop -- a dynamic answer to a dynamic problem

I recently started using Loop, a hybrid closed-loop artificial pancreas (AP) system. I encountered some start-up difficulties with two consecutive infusion site failures and miserable numbers. I put in another style infusion set yesterday and my metabolic world changed. I woke up this morning to this overnight CGM trace:

yellow line=140 mg/dL, red line=65 mg/dL
Time in range = 100%, time hypo = 0%, standard deviation = 8 mg/dL, average BG = 103 mg/dL

I have had success approaching this performance without an AP but these kind of lines are hard for me to repeat. Consistency is a big challenge. Loop made dozens of decisions during the night moving my basal rates up and down. All I did was sleep! I’m thinking that Loop can provide that missing ingredient, consistency.

I have struggled with T1D for almost 33 years. It has taken a bite out of my life’s energy that I would have never given up willingly. This experience rises well above any of the diabetes technology that I have tried over the years. It engenders hope to the extent I had not thought possible. Outside of an outright cure, this is the best thing to ever appear in my diabetes world.

Unfortunately, the AP is not commercially available but the earliest commercial versions are only about six months away from introduction. Ambitious experimenters may find help in various online communities.

I’ve heard and read many hopeful treatment ideas during my 33 years of living with this beast. They sound promising and credible but until today, each and every one of them lingered at that ephemeral five-years distant horizon. Smart insulin and encapsulated beta cells are two that attracted my attention. The artificial pancreas, however, is here today!


Most impressive.

What is Loop comprised of? What parts are involved in this setup? Thanks.

Loop uses four main components. The brains of the system reside in the Loop app on an iPhone. I use the iPhone 6. The most common pumps used are MiniMeds circa 2003-2010. I’m using the 722 model. Loop employs a Dexcom CGM – mine is the G4 with Share. Finally, Loop communicates through a small wireless device called RileyLink. RileyLink is fluent in MiniMed RF frequencies for the pump as well as Bluetooth Low Energy frequencies used to communicate with the CGM and iPhone.

Loop app dashboard on my iPhone

The MiniMed 722 pump


Dexcom G4 with Share


Are you using this throughout the day, or just when you sleep?

I’m now using it around the clock. I’m learning about carb entries and how to deliver insulin from the iPhone app itself. I just discovered I can also enter carbs and deliver a bolus from my Apple Watch.


Woohoo! So happy you got this working, @Terry4.
Hope you enjoy many more nights of uninterrupted sleep!
What is the insulin active showing: the IOB or the actual amount of insulin that is in the bloodstream?

1 Like

Not sure. I think it’s the meal insulin plus the basal insulin but your question will send me looking for the answer.

Active Insulin is the IOB, which includes bolus and any temp basal effects.


Where can we get this Loop app? Also, why does it show you going to 29 mg/dL ? Yikes!:astonished:

@Anthony, so I’m not sure, but at least for openAPS that’s usually a *forecast of where you’ll be in a few hours if the current trend continues.

Tia is right. It’s an early forecast that changes a lot as time goes by. I think of it like weather forecasts; the 7-day forecast is not as reliable as tomorrow’s forecast. Weather can be so unpredictable – probably a poor metaphor to choose! But you get the idea.

That forecast was made at 9:23 a.m. for a 2:00 p.m. BG. Here’s the updated screen at 1:20 p.m.

That’s exactly what it is. The thick “Glucose” line shows actual BG readings up to now. The dashed line shows Loop’s projection several hours (DIA, to be precise) in the future. If the trend continued as predicted (which is by no means guaranteed), @Terry would have had plenty of time (hours) available to make corrections.

1 Like

“The economy depends on economists to the same degree that the weather depends on meteorologists.”


I’m curious as to how these loop algorithms work. For example, if you bolus for a meal at what point does the algorithm then try to begin correcting? How does it know how long it takes the whole meal to digest? Is there some new carbs-in-system value similar to IOB?

So, say you under bolus for a meal, then BG begin to rise and rise, when does the correction algorithm kick in?

I am not totally conversant in Loop terminology and definitions but it’s my understanding that Loop considers all its inputs, BG (blood glucose), Insulin dosed, IOB (insulin on board or active insulin), DIA (duration of insulin action), Carbs eaten and then decides whether to put on the brakes, coast, or speed up.

I think it does this every five minutes so that it revisits previous decisions and updates their execution. The more I learn about it the more fascinating it is. If the Loop sees that it needs to put on the insulin brakes, it can totally turn off basal insulin and pull down IOB. And it does this sooner than I would as a human running in manual mode. Loop sets temporary basal rates for 30 minutes but it can update that plan every five minutes.

It is available at a developer’s repository know as GitHub. Here’s the link:

This is all still novel to me and I’m just learning my way around. I ended up getting an Apple Developer’s certificate that granted me the privilege to pay Apple $99/year and be able to use this code without the necessity to renew it every seven days.

1 Like

I don’t see how this can really ever be truly closed-loop because we will always need to enter in how many carbs we are eating (or what size meal we are eating (as in the Bigfoot loop)).

It really needs to know somehow from the patient how much and when we are eating. This is still a great advancement but I don’t see how this will be overcome, even in the distant future.

It’s all about Inputs! As @Terry4 mentioned.

If my attention and blood sugar decisions are limited mostly to announcing meals to an AP, I’m seeing this as a huge quality-of-life boost.

I can just see the not too distant future, “Hey Echo, start my breakfast insulin routine for eating in 60 minutes.” :wink:


this is soooo amazing!!! i cant even believe it! :open_mouth:


@Anthony, so in theory, using the existing paradigm of basal/bolus it may not be possible to eat, say 120 grams of carbs and not bolus at all. But with some fiddling with parameters in our model lately, I actually have confidence it will be possible to get, say, an A1C of 6.5 with no advance notice on eating. IF you want an A1C of 5.5, that may be tricky without even announcing a meal, but I think it actually might be possible if you log a rough guesstimate of how many carbs you have eaten.
We’ve found that if you toss out the notion that the baseline basal you program into your pump is equal to the amount you need to keep your blood sugar level and instead think of the variables as essentially black-box parameters that tune the algorithm, we often can just let the openAPS auto-correct for carbs our son has eaten as a snack, even without logging them. And while it doesn’t work every time, there are many, many days where our son can eat lunch and I don’t even have to bolus; I just log the carbs on my phone remotely and let the program bring him down. He usually only spikes to about 120 or 130 that way.

1 Like