Hybrid Closed Loops

First a couple terms as we embark on this discussion:

Hybrid closed loop (HCL): An automated system that modulates a pump’s insulin delivery, based on a connected CGM, to try to keep user at a specified target BG. The “hybrid” part of the description means that the user is still responsible for bolusing food themselves in the traditional sense (carb counting and a carb ratio). The term “artificial pancreas” is often used interchangeably with HCL. For the purposes of this post, I’m going to shorten the term to simply “closed loop”.

Automatic bolus: The first closed loops only modulated insulin delivery by settings temporary decreases/increases in basal insulin deliveries. Some closed loops are also capable of delivering boluses in some situations to help bring high blood sugar back to target a bit quicker than basal modulations usually achieve alone. The term “automatic bolus” does not refer to the bolus involved with meals.

At a recent diabetes conference, there was a lot of buzz around the options for closed loop systems. At this conference, I had the privilege of speaking about my perspective of the JDRF Open Protocol Initiative. If you haven’t heard about it, you can read more about it here. In a nutshell, there’s a shift in the industry leading to potentially more choice for us (the people living with diabetes) in the near future. The shift means that you’d be able to pick your favorite CGM, favorite pump, and favorite looping algorithm/controller to combine together into your own little “customized” closed loop. So, instead of all the pieces being pre-picked for you…you get choice about the individual pieces and they don’t all have to be from the same company. The term describing this mix-and-match idea is “interoperable” components. The Open Protocol Initiative seeks to clear the regulatory, legal, and funding pathways needed get interoperable devices on the market and FDA approved.

Interoperable’s Current Status

There’s three components that make up the necessary part of a closed loop; CGM, insulin pump, and the controller/algorithm. In order for a potential interoperable device to be sold in the US, it will need FDA approval specific to that category…and the nicknames used for these categories early on were iCGM, iPump, and iController. To borrow Tandem’s wording, “This new category of devices will make it easier for separate companies to integrate their products into advanced automated insulin delivery systems without having to resubmit each of the components and their associated clinical data every time. It’s a great move on the behalf of the FDA to make it easier for companies with leading-edge technologies to more quickly develop and deliver innovations for the diabetes community.”

So what’s the status on any devices getting approvals for these new categories?

iCGM: The only iCGM thus far is the Dexcom G6 system. The FreeStyle Libre 2 has been submitted for FDA review for this iCGM designation…and it’s taking a little longer than initially expected.

iPump (ACE pump): Ultimately when the first device was approved, the official name of the iPump category was made “ACE pump” standing for “alternate controller enabled”…rather than iPump. The first ACE pump  was Tandem’s t:slim X2 pump. Insulet’s DASH pod were the next to get ACE approval. (FAQ answer: the Medtronic 670G pump is a pump that does closed loop…but it is NOT an interoperable pump. The pump was approved as part of a defined system, meaning it must be used with the Guardian 3 sensor and the algorithm inside the 670G pump itself in order to access closed loop operations…so the 670G pump is NOT an ACE pump)

iController (iAGC): Ultimately when the first device was approved in this category, the name became “iAGC” standing for “interoperable autmated glycemic controller). The recent approval of Tandem’s Control-IQ as an iController marked the first (and so far only) for the category. Tidepool announced its intention to take a version of Loop app through that same approval process and is hopefully close to submitting to FDA soon.

Development Partnerships

Partnerships have been, and continued to be, announced all the time as part of this industry movement…a signal of which devices may be available in the future. It is nice to see that there are more partnerships that appear to be giving more choices for the pairing of system components.


With all this awesome discussion about choice…some people could be in a little bit of a pickle right now in terms of finding the proper information so that they can choose the best system for their needs/wants. So, let’s take a quick peek at the currently available commercial systems, DIY systems, and upcoming systems in trials now. But, I’m adding some color commentary on the systems based on the feedback I’ve heard from other users, too.


There are two commercial options right now; Medtronic 670G and Tandem’s Control-IQ.

Medtronic 670G

Medtronic 670G overview: This was the first commercially available option, approved in September 2016 by the FDA. The 670G uses the 670G pump and Medtronic’s Guardian 3 CGM. When using the closed loop feature for the 670G, it is called being in “auto mode”. The standard target setting is 120 mg/dL and the target can be set temporarily to 150 mg/dL for exercise and other events. The system has you use it in “manual mode” for at least 2-3 days before you go into auto mode. Although you have a scheduled basal profile that you enter into the pump and will be used in manual mode, things change with auto mode. In auto mode, the 670G uses a basal schedule that it decides is appropriate from the previous 6 day’s experiences, not the one you’ve programmed into your personal profile. You will not be able to have a corrective bolus recommended for a high blood sugar until you are above 150 mg/dL and then the corrective bolus recommendation will only provide a recommendation to reach 150 mg/dL. These recommended corrections are not applied automatically, the user will need to implement them manually.

670G Clinical data: To read up on the pivotal trials data that lead to the approval of the system, diaTribe has an excellent summary article here.

I also found this data from this study of the real-world users of 670G after launch to be particularly interesting. The interesting parts?  Teens have the most trouble, staying in auto mode about 66% of the time and achieving a 62% time in range (70-180 mg/dL).

670G user experiences: There’s a whole world of experience that needs to be considered beyond what a clinical trial would tell a user. Beyond time in range and A1C is an evaluation of how easily you can LIVE with the requirements of using the gear. Here’s some of the most frequent user experiences I’ve seen…

  • You cannot change the 120 mg/dL target. No going lower.
  • You cannot use a different CGM (such as Dexcom) and still use “auto mode”.
  • Users have often reported setting very aggressive settings changes in order to achieve better time-in-range. The only two that you can change in auto mode are the duration of insulin action (DIA) and carb ratio. You do not have the ability to simply adjust basals to make the system more aggressive, because the “auto mode” operations are automatically setting your basal profile based on the previous 6 days of data.
  • You must calibrate the Guardian sensor twice daily, sometimes more often. The sensor will provide additional prompts for calibrations when it is unsure of the accuracy of its readings. This leads to what’s called a “pretty needy sensor” experience. It is often cited as one of the major reasons that the 670G auto mode has drop-out rates from 33-49% of its users in studies…they just don’t like all the work required to stay in auto mode and intentionally choose to stay in manual mode. For those users that try to stay in auto mode, there is still a low percentage of time in actual auto mode consistent with the clinical trial results.
  • There are no “automatic bolus-type” corrections…all corrections are accomplished through basal modulations.
  • If you stay at “too low” or “too high” of a basal rate for too long, you’ll be kicked out of auto mode. Practically speaking this means the times when you’d most want to be in auto mode critically (stuck on lows or stuck on highs), you’ll be kicked back to basals in you entered in your profile settings within 2.5-4 hours (see below) of these limits. *Note, Medtronic does not tell you the value of your “personal minimum delivery rate” and “personal maximum delivery rate”. It’s a black box mystery.
  • You will be forcibly kicked out of auto mode, and placed in a “safe basal” mode for 90 minutes (no automated basal adjustments will happen in safe basal mode) if any of the following issues below occur. If the issue is not resolved during that 90 minute span, the user is dropped back into manual mode and will not be getting automated basal adjustments:
    • CGM is outside of the 40-400 mg/dL range
    • CGM has not been calibrated within 12 hours
    • Auto Mode has been at your personal minimum delivery limit for 2.5 hours
    • Auto Mode has been at your personal maximum delivery limit for 4 hours
    • Auto Mode detects that your sensor might be underreading
    • An entered BG is 35% or more different than your current CGM value
    • No CGM data has been received for more than 5 minutes
Tandem Control-IQ

Tandem Control-IQ overview: The Tandem t:slim x2 pump combined with a Dexcom G6 sensor can be used with Control-IQ algorithm to provide a closed loop. This system has been on the market since early 2020. Previously, many users were using the low-suspend-only version called Basal-IQ. The standard target setting is 112.5-160 mg/dL, but there are two alternate target settings users can implement. Sleep targets are 112.5-120 mg/dL and exercise targets are 140-160 mg/dL. With the standard targets enabled, the behavior of Control-IQ can be summarized as shown based on the predicted CGM in the next 30 minutes:

The system uses an initial weight and total daily insulin entry, when entering closed loop mode on Control-IQ, to help establish its model and maximum settings. You don’t get dropped out of closed loop automatically by any condition except CGM data not being refreshed within 20 minutes.  If that happens, you’ll be reverted back to your saved profile settings and no automated insulin delivery changes will happen. As soon as CGM data becomes available again, the system will automatically resume Control-IQ operations and you’ll be back to closed loop operations. There are automatic boluses available, if you are using standard targets. You will get 60% of the needed correction to 110 mg/dL as an automated bolus, when BGs are predicted to go above 180 mg/dL in the next 30 minutes. You can get one automated bolus no more frequently than once per 60 minutes since a previous bolus event (either automatic or manual bolus).

Tandem Control-IQ clinical data: To read up on the pivotal trials data that lead to the approval of Control-IQ, diaTribe has an excellent summary article here. One notable take-away when I initially read this article was that all users stayed in the trial and stayed in closed loop mode 92% of the time. This was a significant improvement from the first-to-market 670G’s results. Additionally, Control-IQ achieved an improved time in range from 59% baseline to 70% after using Control-IQ (670G had clinical trial from 67% baseline to 72% after using auto mode, for comparison). So, while the  data looks like it achieved similar time in range ultimately, I’m actually more impressed at the difference in improvement…Control-IQ brought people up from a far more challenging starting point than 670G did.

There aren’t any real-world clinical studies yet of Control-IQ users, but the Basal-IQ system ended up out-performing the pivotal trials data. I’m excited to see how Control-IQ performs.

Tandem Control-IQ user experiences: To be fair, Control-IQ has not been on the market for nearly as long as the 670G has been…so user experiences are still being gathered. Thus far, the results have been very positive. Complaints that I’ve heard include the usual (can’t aim for lower than 112.5-160 mg/dL) and mundane (pump clip sucks). There have been reports of some users keeping sleep targets turned on 24/7 to achieve a lower overall target range throughout the daytime. Users have been requesting to see more data about their Control-IQ status and to have it link to a mobile phone app (and that is in the works, but I don’t know of a launch estimate on that). There are a couple real-world reviews from users here and here.

An interesting “interoperability” note…this platform could conceivably be the first on-market actual interoperable choices since the announcement of a development partnership between Abbott and Tandem in October 2019. If Abbott’s Libre 2 gets approved for iCGM status (currently under review by FDA), then Tandem customers, using the t:slim x2 ACE pump and the Control-IQ iAGC, could have the G6 or Libre 2 as their iCGM choices. There’s still work to be done for sure to make that happen for Tandem, but it’s an exciting pathway to see finally achieved.

My daughter has been using Control-IQ for about 2 weeks now (after nearly 3+ years using Loop and half a year of OpenAPS). The results so far have been very encouraging. I’m going to do a separate blog post about that when we hit the 30-days of use so that I can have more comprehensive data to share. I never trust that diabetes isn’t just playing nice every time she tries a new system, so I am waiting for more data to really believe it. Thus far, it is out-performing our Loop experience though.


There are many DIY closed loop systems, but the most notable three are Loop, OpenAPS, and AndroidAPS (each of those are links to read more about the individual systems). Most of you know me from Loop, but we also used OpenAPS for quite sometime as well. We have never used AndroidAPS because (1) we don’t use android phones and (2) we don’t have access to the pumps that are compatible with AndroidAPS (those are more commonly available in Europe).

DIY systems are not FDA approved, nor do they come “pre-assembled”. You will have to do some work yourself to build the system and understand how to operate it. There’s a great comparison between each of the systems here. But the short summary is that picking which system is right for you will depend on what aspects you value the most. For example if you absolutely won’t use an Android device, then you need Loop or OpenAPS. Pump choice will also make a large difference. For example if you won’t move away from using your Omnipods, then you need to consider Loop (although AndroidAPS is getting close to a stable inclusion for Omnipod). AndroidAPS is the only system that has the ability to remove the use of an in-between device (such as a RileyLink or Edison rig) because it supports some versions of pumps that have built-in bluetooth capability. AndroidAPS also has the ability to send remote boluses through text messages (although that concerns me a bit from a safety aspect, but that’s a larger topic for another discussion).

All of these systems allow for a level of transparency that is currently unavailable in commercial systems. You can see every aspect of the algorithm’s decision making by examining the code itself and the documentation online. As well, they all integrate with Nightscout and upload live-time with CGM, carbs, insulin, temp basals, and predicted BG values. This live-time view gives remote care givers incredible insight into every aspect of the looper’s status.

There’s also some drawbacks of using a DIY system. You won’t be able to just pick up the phone and call customer support if you have a question (although…holler to the Facebook Looped group for always having pretty quick answers). You may find resistance from your health care provider or school system for using a not FDA approved system. You will likely need to rebuild your system periodically as the code or OS systems of devices get updated.


Tidepool Loop: Tidepool is working on developing a version of the DIY Loop system as an iAGC (that will not require a RileyLink). So far, Tidepool has shared that it is partnering with Omnipod and Medtronic for the ACE pump and Dexcom G6 for the iCGM. They have already stated that the initial platform will be on an Apple device (Android devices will follow later). The observational study has been going for 12 months now and is due to collect data through March 31st. Jaeb Center released initial results at the ATTD conference on February 20, 2020. You can download that presentation here or see it in the browser here. The results are encouraging, with time in range increasing from baseline 67% to 73%. However, as noted in the study, the starting user was demographically from an already “above-average” population. The teen population rose from a baseline 63% to 68% time in range. The time in closed loop mode was 79% on average. Generally speaking, Loop does not go out of closed loop mode unless CGM data is missing, RileyLink communications are failing, or the user has chosen to manually end closed loop operations for awhile. It is unclear from the data so far why the time in closed loop mode is relatively low given the simple requirements for staying in closed loop. No word on the timing of when Tidepool Loop will be available, they are working on getting a submittal ready to send to FDA as of the date this was posted. We also don’t know the specifics on what will be different from DIY Loop ultimately, but Tidepool has stated they are trying to stay fairly similar to the app that DIY users have been familiar with so far.

Note: Tidepool has stated often that it wants to partner with more interoperable devices to provide more choice. So far, Tidepool Loop has partnered with Medtronic and Insulet ACE pumps…but if other companies develop ACE pumps, they could conceivably partner with Tidepool Loop, too. Same for iCGMs, currently just Dexcom G6 but they want more partners to promote more choice for users.

Insulet’s Horizon: Insulet filled its enrollment need for pivotal trials recently and that trial is on-going. They previously conducted a 48-72 hour study with 14 participants and a 36-hour in-clinic study with 58 participants.  What do we know about it right now? Well, the algorithm is contained within the pod itself. So if you walk away from the Android device that controls the pod (aka the PDM), the system will still be able to modify your basals while you are away from the PDM. That’s a pretty useful feature. What we don’t know is any details about the algorithm’s strengths or weaknesses in every day use.  As the 670G and control-IQ experiences have demonstrated, the final user experience will be more than just simply what the time in range or A1C is possible.

Note: Insulet’s Horizon appears to be headed to a partial interoperable product since it has announced partnerships for the Horizon system with Abbott and Dexcom CGMs. Users would need the Horizon system to round out the ACE and iAGC approval portions to be able to switch between CGMs on Horizon.

Beta Bionics: The iLet closed loop system being developed by Beta Bionics has a major difference already…they are aiming for dual-hormone closed loop. So in addition to the usual insulin cartridge in a pump, the iLet will have a pump-friendly version of glucagon to help prevent lows. Beta Bionics is about to begin enrollment for a pivotal trial of their Gen 4 iLet in its insulin-only configuration.  Once they complete a pivotal trial, they should be able continue with the next pivotal trials for dual-hormone looping. Beta Bionics has been around for quite sometime now, and this article has a good history and information:

“The iLet is different from the current commercially available pumps in a few ways: it does not require any settings (carb ratios, basal rates, and correction factors), nor does it require carb count entries for meal boluses (though you do still have to announce meals).

This is an adaptive system that is constantly fine tuning and adjusting its insulin delivery settings, and uses the user’s weight as the starting point to dose insulin. This means that users may expect the pump to go through learning curve at pump initialization with more volatile blood sugars, as it may take a few days for the pump to learn about your specific diabetes patterns and management needs.”

Bigfoot: Bigfoot was originally on-scene when they purchased Asante pumps around 2015. We do know that the Bigfoot smartloop would utilize an app to control the pump and run the algorithm. Bigfoot also has states they intend to use prefilled cartridges and a streamlined ordering process. Bigfoot announced the first enrollments in a clinical trial back in 2016, but to date has not begun pivotal trials to support an FDA submission. Awhile back, Bigfoot raised capital and stated they had hoped to have pivotal trials in 2018, but that schedule apparently slipped as Bigfoot changed CGM partners from an expected Dexcom to Abbott. Unfortunately, there’s not much else that I know about Bigfoot’s plans as it appears their progress is currently quickest on smart-pens.

Leave a Reply

Your email address will not be published.