Disclaimer: I am not a medical professional and everything in this article is based on personal experience and is for informational purposes only.
My son, Jamie (aged 9), is a type 1 diabetic. He was diagnosed 3 years ago, just before his sixth birthday, and ever since then, we’ve been learning everything we can about it, the treatments and the technology that can help. The standard therapy for Type 1 Diabetes (T1D) is multiple daily injections (MDI), where the patient takes two different types of insulin, long-lasting insulin taken once a day and fast-acting insulin taken with meals, so for most patients, this means at least four injections per day.
On top of this, you have to monitor your blood glucose (BG) levels via a finger prick test, both before eating and approximately two hours after eating, so that at least six times a day you have to make yourself bleed on purpose. If at any point you feel “low” or “high” you have to do it again to check to see what your BG level is and potentially have to have another “correction” injection.
This is all a lot to have to deal with, and it is not just for a couple of days or weeks, it is every single day, for the rest of your life, you can’t just take a day off T1D. To make his life easier we started to look about for what was available, how it might help and how we might go about getting it and using it.
Our First Diabetes Tech
The first piece of technology we started using was the Freestyle Libre, which we started using around February 2018. The Libre is a “flash glucose monitor”, which you wear on your arm with a small canular placed under the skin which measures your glucose levels from your interstitial fluid. The glucose level is approximately the same as your blood glucose level, however, it changes slightly slower.
To get a reading from the Libre, you hold a receiver near to the sensor and data is transferred between the sensor and receiver using near field communications (NFC), the same technology used for contactless credit/debit cards. This gives you a current reading plus data for up to 8 previous hours, at 5-minute intervals. Therefore, as long as you scan the sensor at least once every 8 hours, you will have a complete daily picture of the BG levels.
This was good and allowed us to start to visualize Jamie’s BG levels over the course of a day, however, whilst this was nice it didn’t alert us to events such as lows and highs, and we found we were scanning the sensor at regular intervals just to see what it said. Also, for us, the Libre was pretty inaccurate, particularly at the low and high ends of the scale, often telling us he was either hypoglycemic (hypo – low BG) or hyperglycemic (hyper – high BG) when a finger prick test revealed he actually wasn’t. The Libre is factory calibrated and there is no way to calibrate from the patients BG readings.
Moving to a CGM
In July 2018, Dexcom released the latest generation of their continuous glucose monitor (CGM), the G6. The device is similar to the Libre, in that it takes glucose readings from the interstitial fluid, however, it has many advantages over the Libre. Firstly, it actively sends data from the sensor to the receiver, meaning it can automatically alert to low and high events. Secondly, whilst it also comes factory calibrated, it can also be calibrated from the patients BG readings, meaning, if it is “out”, you can give it some additional data to bring it back into an acceptable range. Thirdly, the core algorithm in the device for calculating the BG is more advanced and accurate, meaning it was the first device of its type to be approved by the FDA to bolus from, meaning you didn’t need to finger prick test for making decisions on insulin dosing for meals and corrections. Lastly, it could use a smartphone as a receiver, via Bluetooth, instead of requiring a separate receiver, meaning it can automatically send the data to Dexcom’s servers and the patients family and friends can “follow” the patient and see the patients’ data on their own smartphone and be alerted to events as well. Similar to the Libre, the Dexcom provides a BG reading every 5 minutes.
The price point of the Dexcom G6 was also significantly less than that of the previous generation G5, so we made the decision to move from the Libre to the Dexcom and the difference was night and day for us. It was particularly helpful when Jamie was out of our care, e.g. at school, at friends houses, etc… as we could see his readings remotely and be able to provide assistance to whoever was looking after him if required, often via WhatsApp messages to the same smartphone being used as the receiver.
At this point, we had managed to minimise one part of the T1D treatment regime, the finger prick tests, coming down from an average of 10 per 24 hour period to around 2 per 10 days. The Dexcom sensor transmitter lasts for 10 days (normally) and we always do an initial finger prick test after the 2-hour warm-up period just to check the sensors accuracy and to provide it with at least one calibration value. The Dexcom is normally very accurate when compared to a finger prick BG, so we don’t normally have to finger prick test again, but if the data doesn’t look right, e.g. sudden big jumps in values or a period of values that are highly variable for no reason, we will finger prick test, but this is not very often.
Time to Pump
Next, we started to look into what we could do about the multiple daily injections and had already spoken previously to the hospital about an insulin pump, a device worn by the patient that provides insulin via a canular placed under the skin. The majority of these devices are worn in a “pouch” of some description and link to the canular via a tube. Jamie was adamant he didn’t want this and we also didn’t think this would work for him either, as an active boy. We found out about a “tubeless” pump, the Insulet Omnipod. This is an “all-in-one” pump worn directly on the body and is sometimes referred to as a “patch-pump”, as it is stuck to the patient’s body like a patch. This pump comes with a separate handset, called a Personal Diabetes Manager (PDM), which is used to control the pump.
Pump therapy allows much finer grain control over the patient’s diabetes for two main reasons:
- The pump can give much smaller doses of insulin, in the case of the Omnipod, as small as 0.05 units, whereas with most common insulin pens used for MDI therapy, the smallest possible dose is 0.5 units.
- The patients no longer use a “background” insulin (basal), given all in one dose every 24 hours, which is meant to be absorbed by the body evenly throughout that 24 hour period, but of course, it isn’t, and is different from patient to patient. With pump therapy, the pump gives the patient fast-acting insulin constantly to act as the basal insulin and as such, different amounts of basal insulin can be given at different times, which, if you can get it right, should give the patient a much steadier BG. Also, the basal insulin can be temporarily suspended if needed, e.g. if the patient is going to be more active, e.g. taking part in a sport or other physical activity for a short time.
We spoke to Jamie about the Omnipod, went along to some information sessions at the hospital with him about it and in February 2019 we started on the Omnipod, so now Jamie had two devices connected to him.
Each pod lasts for three days and the used pods are returned to Insulet for recycling. The pod can be worn on the back of the upper arms, stomach, front of the upper legs or the lower back / upper buttocks area. Jamie tends to wear it on his stomach or upper legs and occasionally on his upper arms, but he tends to knock it more when it is on his arms and we have had a couple come off because of that, most significantly once when Jamie was at basketball practice and another child knock into him and “pulled” it off.
So, we had these two pieces of tech on Jamie’s body, which were related but they do not “talk” to each other. The Dexcom would tell us what his BG was and the Omnipod would administer the insulin, via the programmed basal rates and the boluses we would give for food and corrections, but it all required user intervention. We got pretty good at knowing how different foods affected Jamie’s BG in different ways and using the different types of bolus delivery, e.g. all at once, extended over several hours, etc… and by tweaking his basal rates, we managed to get his BG relatively under control, but we would still have regular hypo or hyper events, requiring treatment. Over the past three months, mid-September to mid-December, his average “time-in-range”, the time his BG is between 4 and 10 mmol/L has been 56.6%, which is good, but still not great.
I had heard for some time about “looping” and was aware of a community movement called #wearenotwaiting along with various related projects, including NightScout, OpenAPS and Loop, that were working on creating tools that would connect the CGM and pump together to create a “closed-loop” artificial pancreas system (APS).
We were also already using Tidepool to visualise the data from the CGM and pump in one place and this was an invaluable tool when it came to making decisions to change basal rates, bolus calculation values, etc… and they then announced that they were going to be working on bringing Tidepool Loop, based on the open-source Loop project, to market with Dexcom and Insulet. I decided, therefore, to look into it further and found that we already had most of the required components to use the open-source version of Loop.
So, to set up a closed-loop APS you need some pieces of hardware and some software. These are:
- An insulin pump
- A Continuous Glucose Monitor
- A compatible smartphone, either Android or iOS.
- A RileyLink, to act as the “go-between” between the pump and the CGM.
- Software, this will depend on the setup but either OpenAPS or Loop and optionally Nightscout and Tidepool.
- A computer, for Loop this will need to be an Apple Mac, but for OpenAPS it can be a PC or a Mac.
- An understanding of how to install software on your computer.
- A willingness to try to compile the app from source, if you have never attempted this before.
- A developer account on Google or Apple, depending on which smartphone you are using.
- And time to read the documentation, compile the software and set everything up and then read the documentation again and read the documentation once more.
The part where I think most people would probably come up against a barrier with this is the point at which you have to compile the app yourself, from the source code available via the appropriate GitHub repositories. To a lot of people, I think this would seem like a “step too far” outside their comfort zone, however, I can say at least of Loop that the documentation provided is very very easy to follow and if you do exactly what it says, step-by-step, that most generally computer literate people could manage it.
You might have noticed that I stressed about reading the documentation, and if you go down the Loop route as we have done, you will notice they stress the same. They have spent a lot of time and effort to provide excellent documentation, that covers everything in great detail and explains everything step-by-step, not only for compiling and installing the app but on how to actually set it up and operate it. There is a lot, and I mean A LOT, of information in the documentation, and one read through will not be enough.
You will probably find, like I have, that you are referring back to them time and time again, as you use Loop and begin to understand how it works and you need to make changes to your settings. This is not because the app is hard to use, far from it, it is very user-friendly, but understanding what values to tweak and when is the key to making a closed-looped artificial pancreas system work for you. Despite popular belief, it is not fully automated, artificial intelligence or magic, it is really just a set of calculations based on the data it is given and settings set, that then administers the insulin as it is required by the body to hopefully provide a more consistent BG and more “time-in-range”.
An insulin pump is required for a closed-loop APS and the choice is pretty limited. It is mainly limited to older model Medtronic pumps and the Omnipod “Eros” pod (not the newer Dash version) and then only Loop supports the Omnipod, with OpenAPS only supporting Medtronic pumps. Luckily for us, we were already using the Omnipod “Eros” pump, so from the pump point-of-view, we were good to go.
Continuous Glucose Monitor
A continuous glucose monitor (CGM) is also required so that you have a constant delivery of BG data. The main CGM supported is the Dexcom, from the G4 onwards and the Medtronic Enlite is also supported by both. It is also possible to use Libre with BluCon or MiaoMiao but this is not natively supported, so you need to hook up some additional 3rd party integrations to allow OpenAPS or Loop to receive the CGM data. Again, on this point, we already had the Dexcom G6, so were good to go here as well.
Loop runs on iOS and OpenAPS runs on Android, so depending on the route taken, you will need one or the other. We were already using an iPhone SE as the Dexcom receiver and given Loop requires iOS and only Loop supports Omnipod, we were in luck again.
This is where things get slightly tricky. The RileyLink is the name of the device that allows the smartphone to talk to the pump, acting as a bridge between the two. The Omnipod Eros uses radio signals at 433MHz to communicate with the PDM, so this is what the RileyLink provides. It connects to the smartphone via Bluetooth and to the pump via the same 433MHz radio signal as the PDM uses. The issue, however, is that RileyLink is essentially just a set of open-source schematic’s that you can use as a reference to build it yourself. However, the community comes through here as well, and you can order a pre-built RileyLink board, battery and case from getrileylink.org for US$150, which includes the cost of shipping to the UK. All you have to do is put it together, which is as simple as plugging in the battery and sliding it into the case.
We placed our order via the website and it was delivered 9 days later, so now we had all the hardware components required, it was time to start looking at the software required.
In our case, Loop is the main piece of software required as it is what controls our closed-loop APS. As I said above, this is probably the point at which a lot of people come up against a barrier. At this time, there is no particular way around this, you have to download the source code and compile the app yourself. For Loop, it also requires an Apple Developer account and if you don’t pay the $99 per year fee, the app you compile and install on your phone will only work for 7 days, which would be particularly annoying. With the paid developer account, the app will work for a year, which might sound slightly annoying as well, but you will probably need/want to update it before then anyway and each time you recompile and install, it resets the year countdown, so you will probably never come up against that as an issue anyway.
As a software developer, I already had most of the required software installed on my computer, so I went ahead and downloaded the source code from Github and followed the instructions in the documentation to get the app compiled and installed on the iPhone, which I managed without any issues. It really is pretty straight forward and the documentation explains it all very well, so even if you have no experience of doing this sort of thing, you will be able to do it.
Nightscout is not required but is extremely useful, so is recommended. Setting up Nightscout is not completely straightforward either, as it requires you to provision your own server on either Heroku or Microsoft Azure. Again, the documentation for Nightscout is very good and I simply followed the guide and had it up and running relatively quickly. There is no cost involved with this, as everything you will need is included in Heroku’s free tier.
We had been using Nightscout to view the CGM data for a while. However, once you are using Loop there are some additional settings you can add to your Nightscout configuration to give you Loop specific information, again these are all explained in the Loop documentation here.
Nightscout is great for “in the moment” viewing of the Loop system as Loop works and shows you lots of helpful information:
Nightscout also has a reporting tool that you can use to produce a range of reports from the data it has collected from the CGM, Pump and Loop app. There is also a Loop specific report, called Loopalyzer, to help see what Loop has been doing and if there are any adjustments you need to make. Whilst these reports provide a good overview of the data, they are not the best visually.
Data visualisation is where Tidepool excels and it is much easier to set up than the other software components. Simply register for an account, download the app from the app store, log in to the account on the app and authorise it to access your Apple Health data, it will then take care of the rest for you.
We had actually been using Tidepool to visualise the CGM data almost since we first started using the Dexcom. It also has an uploader tool that allows you to upload data from other devices, like the Omnipod PDM, so we had also been doing that since we started on the Omnipod in February. It has four main views for the data, these are:
- Basics – Which give you an overview of the data for a specific date range, showing you BG readings, bolusing, infusion site changes and basals, as well as an overview panel with daily averages, average glucose value, average daily total insulin, average daily carbs and a glucose management indicator (GMI) which is approximately what your A1C level will be.
- Daily – This gives you charts of your BG levels, Bolus and Carbohydrates and Basal Rates for a particular day and you can scroll back and forwards through the dates. There is also an overview panel with time in range, average glucose value, total insulin (split between basal and bolus), total carbs, standard deviation and coefficient of variation (CV).
- BG Log – Shows a chart of BG log events, which only works for BG values linked from the PDM, but as we are no longer using it, it doesn’t show us any data anymore.
- Trends – Gives you a chart of BG over a specified date range averaged down into a half an hour period for the whole 24 hours and then shows you median, 50%, 80% and 100% reading ranges. This sounds very complex, but it actually gives you a really good idea of how your BG has been over the 1, 2 or 4 week period you are viewing and really helps you spot areas that need some work. The page also has an overview panel with similar data to the panel on the Basics page. You can also hover over the chart to reveal more of the underlying data so you can analyse further. This report is really where Tidepool comes into its own.
Tidepool also has a secondary feature, that is not directly useful to you as the end-user but could be highly useful to the T1D community as a whole, it allows you to opt-in to donating your de-identified device data, securely and anonymously to researchers, device makers, and other innovators who can use it to develop and improve diabetes products. You can even pick and choose which projects your data can go to. Personally, I just selected them all because, hey, why not. For more information about this check out the blog post about Tidepool’s Big Data Project.
So, how has it been?
We started off with the settings from the PDM, but we knew they would probably not be the best and we would need to tweak them, but we are only 5 days in and it looks so good already. One thing that we noticed right from the first night, was that Jamie’s overnight BG values, which have always been an issue for us, were so much better. Jamie had always suffered from either excessive highs or lows during the night, with no two nights being the same and despite changing settings, trying to use temp basal rates, etc… we rarely managed to get through a whole night without his Dexcom alerting to either a high or low. However, the last 4 nights have been completely different, not a high or low insight, almost completely flat and in range throughout the night and we have actually managed to get some uninterrupted sleep.
Even if this is all Loop gave us, it would have all been worth it, but we have been actively tweaking his settings for the first 4 days and it looks like we have started to find a setup that is working during the day as well. Yesterdays highest BG value was 11.8 and today’s was 13.1. Also, we have had no hypos at all since starting on Loop. To say we are pleased with the results we have seen so far would be an understatement.
It has also given us some additional reassurance that something is constantly watching over Jamie and unlike the “watching” provided by the Dexcom alone, which is already reassuring as a parent, Loop is actively changing his insulin dosing to keep him in range and helping to avoid lows and highs.
We are so thankful to everyone in the #WeAreNotWaiting community who have put all this amazing technology together and provided such an amazing system, giving their own time, knowledge and effort for nothing in return other than the thanks of the community. We salute you!
Here is a list of links to resources, apps, etc… we have found helpful during this journey. They are in no particular order.