Understanding your options and how they compare can help you decide which hybrid closed loop system would best meet your needs. Not your doctor’s. Not your health insurance’s. YOURS.
There are several HCL systems available: some that have been approved by government authorities and are produced/marketed by major diabetes device companies, and others that are classified as “DIY” (Do It Yourself) – systems that the user builds on their own using commercially-available components. Down the road, this will likely change radically as “interoperability” comes to the forefront. With interoperability, users may select a pump, CGM and algorithm from different manufacturers and have them work together seamlessly. Until then, the systems are mostly integrated by manufacturer.
Hybrid Closed Loop Systems Options
The current commercially available systems include the Medtronic 670G and Tandem X2 with Control IQ.
Medtronic was the first company to launch an HCL back in 2018. It features the Medtronic 670 tubed insulin pump and the Medtronic Guardian 3 sensor. The sensor lasts for seven days and requires a minimum of two fingerstick calibrations daily. It uses a rechargeable transmitter which communicates with the pump using radio signals. Within the pump is an algorithm called “auto mode” that adjusts the pump’s basal insulin delivery every five minutes based on the sensor readings. The algorithm starts with a flat 24-hour basal program (based on the user’s total daily insulin requirements) and adjusts the basal rate from there. There are set limits to how much (and for how long) the algorithm will raise or lower the basal delivery. The system uses a preset target of 120 mg/dl when adjusting basal delivery and a target of 150 mg/dl when calculating correction boluses. A temporary basal adjustment target of 150 mg/dl can be set prior to exercise.
Tandem’s Control IQ system features the sleek Tandem t:slim tubed insulin pump with a full-color touch-screen.
CGM data reaches the pump via Bluetooth signals from the Dexcom G6 sensor, which lasts 10 days and requires no calibrations. Activating the Control IQ algorithm requires the mere flip of a switch and entry of the user’s weight and average daily insulin requirements. The algorithm predicts where the glucose will be in 30 minutes (based on CGM values and trends) and adjusts the pump’s usual basal delivery based on the predicted value.
The system can also administer conservative correction boluses on an hourly basis if the glucose is predicted to be quite high or remains high. It has a preset target of 110 mg/dl. Extended boluses (but not temporary basal rates) remain an option when Control IQ is running. The system has two optional “activity” overrides: An exercise setting (which makes the algorithm more conservative) and a sleep setting (which makes the algorithm slightly more aggressive but turns off automated correction boluses).
The current “build-it-yourself” systems include the Loop app, OpenAPS and AndroidAPS
All three consist of a combination of FDA-approved products (the pump and CGM) and a non-approved app which serves as the controller/algorithm. The apps were designed and built by people in the diabetes community who were looking for a better system than those that are commercially available. The apps themselves are free; they can be built by following a series of instructions that are available online. However, they do require the builder to have a software developer license (specific to Loop and available to anyone for a modest annual fee) and some ancillary parts for allowing proper data transmission between devices (once again, for a modest fee). Because the Loop app sits on the user’s phone, most pump programming (including bolus calculations and programming) can be done right on the phone.
Loop can interact with older-model Medtronic Paradigm tubed insulin pumps (firmware 2.4 or lower) and previous-generation (Eros) OmniPods – both of which transmit and receive radio signals. It can receive data from the Dexcom G5 or G6 CGM. The app can be built on an Apple computer and runs on an iPhone or iPod Touch. It requires the use of a small signal converter called a “RileyLink” which can be ordered online for approximately $150. Loop applies the user’s customary basal rates as a starting point, and adjusts basal delivery based projected glucose values (from sensor data), active insulin, and active carbs. It allows the user to choose a maximum allowable basal level as well as the degree of adjustment whenever predicted and actual glucose levels vary. It applies the user’s preferred carb ratios, correction factors, targets, and insulin-on-board algorithms. Although extended and combination (square/dual) boluses are not available, Loop offers the option of entering the “absorption time” for any carb entry. With a longer absorption time, less bolus will be delivered up-front, and the basal delivery will be increased over the next several hours. Finally, Loop allows the user to set a variety of “temporary overrides” which essentially make the algorithm more or less aggressive for a chosen period of time. Details about the Loop app can be found at https://loopkit.github.io/loopdocs/
OpenAPS and AndroidAPS are similar to Loop
They can work with older-model Medtronic pumps, and AndroidAPS works with some select older Medtronic pumps and the Accu-Chek Combo and Dana pumps. They are compatible with a variety of CGM systems, including Dexcom G4, G5 and G6, as well as the Medtronic Guardian and Freesytle Libre. OpenAPS runs on both iPhone and Android devices. AndroidAPS runs on Google Android phones. For OpenAPS, communication between the app, CGM and pump requires ongoing web access. This means carrying a small computer circuit board and portable battery. Building the app takes a lot more “coding” and tech know-how than Loop, so it is used mainly by those who are wed to an Android device. Details about OpenAPS can be found at https://openaps.org/. AndroidAPS uses OpenAPS code, but is unique in that it doesn’t require the small computer circuit board to work. Details about AndroidAPS can be found at https://androidaps.readthedocs.io/en/latest/EN/
Hybrid Closed Loop Systems Comparisons
Which Hybrid Closed Loop will take home the award for best supporting system?
At last, the moment we’ve all been waiting for. Which HCL will take home the award for best supporting system? As with a movie, beauty is in the eye of the beholder. Different features are simply more important to different people. Personally, I couldn’t care less about whether or not the government deems my system to be safe. All that matters to me is that it works. Others consider simplicity and convenience to be essential, or customization to have the utmost importance. Out-of-pocket costs play a major role for some.
First, some general observations. Compared to commercially available systems, DIY systems allow much greater flexibility in terms of setting personal parameters. The software can be updated frequently as new strategies and products become available. For example, within weeks of Novo Nordisk’s Fiasp insulin hitting the market, Loop designers installed and algorithm for calculating insulin-on-board based on a Fiasp action curve. Because the DIY apps are interoperable, users have multiple pumps and CGM systems from which to choose. And because the algorithms can be set to the user’s desired intensity level, the outcomes consistently outperform those of the commercial systems.
Of course, there are drawbacks to using a DIY system. They often require the use of older-model insulin pumps. There is no manual, warranty or customer support offered with the app – just an online community of peers to assist with troubleshooting. Users must carry extra equipment to allow the app to communicate with the pump, and the equipment usually requires daily charging. Most healthcare providers have limited experience with DIY systems, and hence cannot offer much in the way of guidance. Some will actually decline to care for patients who choose to use a “non-approved” system. There are limited options for accessing data and generating reports from DIY systems: Tidepool is one option, along with Nightscout. Unfortunately, neither of these integrate easily into the medical records systems at most diabetes clinics.
But perhaps the most significant obstacle to use of a DIY system is the fact that the user needs to “build” the app on their own. Building an app can be a long and arduous process, especially for those with limited computer tech skills. It also requires specific types of operating systems on both the computer and the smartphone.
So here are our ratings (along with some explanation, where appropriate) on the factors of: Safety (minimizing hypoglycemia and system errors), Performance (degree of glucose control that can be achieved), Simplicity (ease of initial setup and daily use), Convenience (effort/disruption incurred by the user on a daily basis) and Cost (for hardware and disposables), for each HCL system.
Give credit where credit is due. Medtronic was the first company to navigate the FDA maze and bring an HCL to market.
In fact, 670G seems to have been designed with the FDA in mind. Everything about it screams “safety first,” from the conservative glucose targets to the SafeMode temporary default feature. Medtronic has consolidated everything into just two parts — the pump and the sensor. The pump contains the insulin adjustment algorithm, and communication between the pump and sensor is virtually seamless. 670G comes fully functional right out of the box (actually, 11 different boxes). There is nothing to “build.” Medtronic provides full training and extensive support for all users and provides follow-up guidance as needed. Although it is a costly system (about $7k up-front plus around $3k annually for disposables), it is covered by most private health insurance plans. But the advantages pretty much end there.
Medicare still does not cover 670G because the sensors have yet to be cleared as a replacement for fingersticks. The sensors themselves are complex to insert and tape in-place. They lag the competition in terms of accuracy, longevity and signal reliability. The transmitters require regular re-charging, and the sensors require fingerstick calibrations multiple times daily.
The pump is relatively bulky and awkward to wear (the clip is even upside-down). The user interface defies most laws of logic. It has menus upon menus upon menus, and requires inordinate button presses to perform even the most basic programming.
Due to the nature of the pump, sensor and “auto mode” algorithm, most users require three or more training sessions and must stay in “manual mode” for about a week before turning the “auto-mode” feature on. When using auto mode, users may not extend or adjust the bolus doses calculated by the pump. Planning to exercise? Tough cookies. Sensor shows a distinct rise or fall at the time of the bolus? Too bad. Take the calculated dose or take nothing at all. To alter the dose, users are forced to “lie” to the system by intentionally entering incorrect carb amounts or glucose values. Correction doses in “auto mode” are also determined by the system, not the user. And the system does not target a glucose of 120 (which is the fixed target for the basal self-adjustment algorithm). It targets 150. Why? Most likely to appease the FDA. The system also does not base correction doses on the “sensitivity” or “correction factor” that the user (and their healthcare team) may know to work, but on a mystery-clad secret sauce that Medtronic refuses to divulge. In fact, the only variables that the user has any control over are the insulin-to-carb ratios and the duration of insulin action. Even the user’s usual basal settings are replaced by a flat 24-hour pattern that is adjusted by the algorithm. For those who require significant basal “peaks” and “valleys” to their daily basal rates, the algorithm will be forced to react to rising/falling blood sugars rather than preventing them in the first place. There are also conservative limits to how much (and for how long) the system will raise/lower basal delivery. Exceeding these limits triggers alarms and takes the user out of “auto mode.” Many users develop alarm fatigue with this system due to the frequent daily interruptions imposed by the pump, sensor and algorithm.
The only allowable adjustment to the algorithm is a temporary target of 150 (rather than 120). Presumably, this is for exercise. Trouble is, it takes a couple of hours for this adjustment to actually get the glucose up to 150, so it has to be set a few hours in advance of exercise. And this still isn’t usually enough to prevent a low with longer and more intense forms of physical activity.
Analyzing data requires an arduous download to Medtronic’s Carelink program. Once considered a cutting-edge program, Carelink is now relatively slow and complex, with reports that are difficult to interpret. Some reports make an attempt to provide analytic insight, but the information fails to take the varied day-to-day lives of the user into account. On the plus side, Carelink is still the only program that generates a true cause-and-effect report for evaluating the effectiveness of the insulin-to-carb ratio settings.
Overall, the 670G takes a considerable amount of work and sacrifice to maintain. Due to the conservative nature of the algorithm, the control that can be achieved is only fair-to-good. For those with very poor control prior to using it, the improvement can be significant. But for those with tight control, the results can be somewhat disappointing.
Tandem X2 with Control IQ:
Tandem built their HCL system largely around the needs of the typical consumer.
The pump is sleek and easy to program, with a display that really pops… but they also screwed up when it comes to the clip (upside down). The upgradeable pump software means that users don’t have to wait until their pump is out of warranty to benefit from new and improved features. The pump links with the simple, reliable, calibration-free Dexcom G6 and uses an algorithm that applies all of the user’s preferred basal and bolus settings… except for the duration of bolus action (which is locked at 5 hours) and the target (which is set at 110). By administering correction boluses on the user’s behalf when glucose levels are notably elevated, the Control IQ algorithm is faster than other HCL systems at fixing high blood sugars. Still, the extended duration of bolus action means that IOB is often over-estimated. This helps with prevention of lows but keeps the user from achieving very tight control.
The algorithm has a couple of override options: One for exercise (which makes the algorithm slightly less aggressive) and one for sleep (which is slightly more aggressive but eliminates automated correction bolusing). As was the case with Medtronic’s “temp target” feature, the exercise algorithm provides “too little too late” for preventing lows with most forms of exercise.
Setting up the HCL algorithm requires nothing more than flipping a switch from off to on and entering one’s weight and average daily insulin use (which can be obtained in the pump’s history). There are no added alarms or user interventions required to keep the system running other than having a steady stream of input from the CGM.
The Tandem pump (along with CGM data) has its own download software (t:connect) which is fast but a bit clunky to navigate. The reports are clear and simple, but little is offered in the way of insight. The pump can also be downloaded to “midware” programs such as Glooko and Tidepool.
Overall, Tandem offers a nice value proposition for users. The glucose-control outcomes are similar to those of Medtronic in terms of average, time-in-range and frequency of hypoglycemia, but the amount of work required by the user to maintain the system is much, much less. It is covered by Medicare and most private health insurance plans.
Necessity truly is the mother of invention. And nobody invents quite like tech gurus in the Silicon Valley.
From a PERFORMANCE perspective, Loop offers many advantages. The target glucose levels are customizable by the user – including a temporary target increase (for exercise) and decrease (for pending meals) to minimize the magnitude of the post-meal peak. Loop allows the user to modify suggested bolus doses for both meals and corrections. And speaking of corrections, Loop applies the user’s customary correction factor and target settings in making dosage recommendations. And then there’s the fan-fav feature: With Loop, you program everything directly on your phone!
Most notable, in my opinion, is the fact that Loop makes proactive adjustments instead of waiting for the blood sugar to start going too high or low. It bases this on active insulin (including bolus and basal insulin) as well as active CARBS – something truly unique to this system. When entering carbs for bolus calculations, Loop asks approximately how long it will take your food to digest (2, 3 and 4 hours are common entries). This is essential for determining whether the insulin-on-board will be offset by the carbs-on-board. All of this is displayed clearly on the app’s status screen, along with a glucose projection covering the next 3+ hours.
When determining basal delivery, Loop utilizes the user’s prior basal settings as a starting point. This is a key feature, because it can prevent overnight and between-meal rises/falls rather than waiting to fix them once they take place. The basal adjustments made by Loop can be much more robust than those made by commercially available systems. The magnitude of the adjustments is limited only by the user’s setup preferences, and there is no limit to the duration of the adjustments.
The customizable nature of Loop and the aggressive nature of the algorithm produces glucose control that is tailored to the user’s preferences… which in most cases means “very tight.”
Loop does a nice job of keeping alarms and alerts to a minimum. It more-or-less runs in the background, and only alerts the user if it has lost the sensor signal or has not been functioning for at least 20 minutes.
Loop currently accepts data from Dexcom G5 or G6. G6 is considered by many to offer the best combination of accuracy and ease-of-use of any CGM on the market. Being a non-regulated product, Loop can be adapted rapidly as newer and better products come to market (essentially as soon as the volunteer software developers can add a new “branch” to the program code). No need to go through months (or years) of product testing and then more months (or years) waiting for the FDA to approve the updates.
Loop can generate reports (Nightscout is the primary program), but you’ll need decent tech skills (or someone with such skills at your side) to set this up and read the reports.
For all its strengths, Loop was never run through vigorous usability testing. It doesn’t come with a nice set of instructions. Or professional training. Or 24/7 phone support. Many of the advanced features in the app are difficult to locate. Those who choose to use an old Medtronic Paradigm pump (as opposed to OmniPod) must endure the outdated nature of the pump and lack of a warranty. If something goes wrong with that pump, you’re basically S.O.P. until you can track down a suitable replacement – and that’s getting harder all the time as the availability of Loop-compatible Medtronic pumps shrivels up.
Then there’s the hassle factor. In addition to wearing a pump and sensor, Loop users must have their app-containing phone handy at all times and carry the RileyLink device to enable communication between components. And yes, the RileyLink requires daily (or nightly) charging. Cost-wise, users will incur a few hundred dollars per year out-of-pocket (for the Riley Link and Apple software developer license) in addition to the usual expenses for pump and CGM supplies. For those who use the OmniPod as their pump, coverage may be available through one’s pharmacy benefits. This can reduce out-of-pocket expenses and may allow users of tubed pumps to switch to OmniPod and begin using Loop even while their existing pump is still in-warranty.
Perhaps the biggest hurdle to Loop use is the initial setup. It is NOT for the technologically challenged. In addition to having the right kind of computer for building the app, you will need many hours, significant patience and attention to detail in order to follow the online instructions for “building” the app and installing it on your iPhone. Even after everything is set up, there are dozens of picky little problems that can keep Loop from functioning properly. Most can be fixed easily by the user, but some may require the assistance of the great and powerful Loop Facebook Group.
Overall, for those who can get past the initial setup, Loop offers superior glucose control with minimal work/effort/sacrifice on the user’s part.
As a DIY HCL system, OpenAPS is very similar to Loop in terms of its design, advantages and drawbacks.
On the plus side, it has an aggressive algorithm and is usable by those who prefer an Android phone. It works with a wider variety of pumps and CGMs and has built-in “auto-sense” and “auto-tune” features for helping the user optimize their insulin delivery settings.
On the downside, because it requires web access for the algorithm to function, there is more equipment to carry around all the time, and there tend to be more interruptions in the automated insulin adjustments. Building the Open APS app is a bit more difficult than building Loop, with more steps and “coding” required. The user interface is also a little less intuitive.