“Making” and “DIY”ing – continued

I had a conversation this week with someone in the CGM in the Cloud Facebook group, after they indicated they wouldn’t be (or maybe weren’t interested in) joining the “dev” group for #OpenAPS – and it’s a conversation I find myself having often. Here’s what I usually end up saying, when someone says they’re not a “dev” or “not an engineer” or something similar:

“I’m not a formally trained developer/coder/engineer, either… but I keep telling people, many people in this project aren’t- it’s a passion project where we learn what we need to learn to do the things we want to do. It’s fine if someone chooses not to do something, but I encourage everyone to not let labels or perceptions of traditional roles stop them from jumping in and giving it a try to see what they can learn and thus do! Especially with this awesome supportive community of people willing to help you as you go.”

This also came up when we were discussing what it takes to be a “maker” on TEDMED’s #GreatChallenges live panel today. One of my excellent fellow panelists (Cole) pointed out that pretty much everyone is a maker – whether you tweak a recipe, work with wood, or find any kind of workaround of any sort to make things work. (Which in my mind makes every single person with diabetes a “maker” and probably anyone with any disease or health care condition that they live with.)

I previously wrote about what it takes to DIY from a DIYPS and #OpenAPS perspective (and why that’s important), but I think it holds true across any aspect of diabetes or any other disease state – and definitely beyond healthcare:

Passion, persistence, and willpower needed.

So please, don’t let labels stop you from DOING. You can learn whatever tech skills you set your mind to. You can find numerous ways to solve a problem, whether it’s on your own or by partnering with someone else – and there’s plenty of people with the skills who are willing to help you learn, too.Remember, we started building #DIYPS to make louder CGM alarms. Scott and I have both learned numerous new things and new programming languages and skills along the way as we went from alarms to an alert and recommendation system to a closed loop artificial pancreas (and now people who own 4 Raspberry Pis). We didn’t come to the table with knowledge of everything we needed to know to do what we first wanted to do – and we’re definitely still learning a dozen or more things (programming languages, new software, etc.) along the way as we continue with #OpenAPS. We also didn’t know anything previously about working directly with the FDA – and now we are, on a number of projects, in order to help scale from n=1 of a DIY artificial pancreas to many n=1s around the world.

You can do this. Bring your passion, and go do great things!

#WeAreNotWaiting, are you?

Why the DIY part of OpenAPS is important

I had the chance to talk about DIYPS and OpenAPS during a demo session in DC last week. (Thank you to Gary from Quantified Self and Marty from the National Academy of Sciences for making this possible!)

I walked away with several insights:

  1. Many people don’t know about diabetes; fewer have a realization of current diabetes tech. In several cases as I was describing the closed loop artificial pancreas, people stopped me and were wowed – but not by the closed loop. They were impressed by the CGM.
  2. Others think that this type of technology is already out on the market.

So, I believe we have a long way to go in communicating and advocating for this type of technology. We know it’s behind where it should be – and we want it to catch up. That’s a big part of the OpenAPS goals to help the FDA, device companies, and everyone involved move a little faster than they might otherwise, because #WeAreNotWaiting.

But here’s the other question I was often asked: “How many people have you given this to?”

I frequently embarked on an explanation of how we can’t “give” away #DIYPS or the OpenAPS implementation – in fact, we can’t and won’t give away the code, either. Some of that is because the FDA says no – and some of it is common sense and principles that both Scott and I hold.

Here’s why I think it is so important to keep the DIY in DIYPS and each OpenAPS implementation that is in progress:

  • You need to have a deep understanding of the system before even considering using it on yourself. You need to know what it’s trying to do in all situations, including the fringe cases (the “this is unlikely to happen but if it does…”), so that you know when it’s working – and when it’s not – whether it’s 3pm in the afternoon at work, or 3am and you wake up and find something is not right and the system is not working.
  • You need to go step by step and test and ensure at each stage that it is working as expected – both in a “this is what it should be doing” and “it is giving out the correct amount of insulin”. Remember, insulin is a lethal drug. It’s also a lifesaving drug. It’s important to remember both of these things and balance the risks accordingly.

From the conversations I’ve had with people interested in learning more or getting a DIYPS-type system for themselves, they fall into two categories:

  1. “How can I buy it from you?”
  2. “What do I need to do to make one?”

Given my above reasoning, the second question is my favorite. The first one scares me, if someone does not then switch to the #2 question. Many people do go from #1 to #2, which is great.

DIYPS, for me, and OpenAPS implementations, for others, are works in progress. They’re not perfect. They’re better than what’s out there (like sleeping through alarms when you’re low at night), but they also have big risks. And it’s important to know, and respect these risks, and understand the limitations of the system, before being able to take advantage of this type of system – and to build the system with appropriate safeguards. (This is one of the reason we have OpenAPS, for example, designed to accept multiple failure points – like walking out of range, loss of connectivity, etc.)

The ability to buy a “black box” type system where you don’t know exactly how it works, but you trust that it works? That will be coming from the major device manufacturers in several years – hopefully sooner rather than later, and that’s something that OpenAPS will hopefully help make happen more quickly.

So to answer the #2 question, what do you need to make a DIYPS or OpenAPS of your own?

I’ll answer the technical aspects of this question in another post, but the first thing I always say is: “The willingness to build and test and test and test some more before ever considering using it on yourself.”

How to do “eating soon” mode – #DIYPS lessons learned

“Do you prebolus for meals with #DIYPS?”

The answer to this question is complicated for me. I don’t “prebolus” like most people do (meaning “take some or all of your meal insulin about 15 minutes before you eat”).

I do take insulin before a meal. In fact, I do it up to an hour before the meal starts, by setting my correction target BG from it’s usual range (usually 100-120) to 80. This usually means I’m usually doing anywhere from .5-1u or more of insulin prior to a meal. But the amount of insulin has no direct relationship with the total amount of carbs I’ll end up eating during the meal.

Does it work? Yes. Do I go low? No, because it is unlikely that I would get anywhere near 80 by the time my carbs kick in for a meal (15 minutes after I eat), and therefore the initial carbs are handled by that initial amount of insulin from the eating soon-bolus. (Last year, I wrote a post about “eating soon mode” under the guise of lessons learned about meal time with #DIYPS – if you want to read the reason behind WHY eating soon mode is key in more detail, you can definitely read the longer version of the post. It also links another key concept I’ve learned about called carbohydrate absorption rate.)

So, how can you manually do “eating soon” mode?

1. If you know you’re going to eat anywhere in the next hour, manually calculate a correction bolus with a target BG of 80. (Example – if your correction ratio is 1:40, and you are currently 120, that means you would give yourself 1u of insulin.) An hour, 45 minutes, 30 minutes – whatever you make work is better than not doing it!

2. Eat your meal and bolus normally, but use your IOB as part of your meal calculation so you don’t forget about that insulin you already have going. (Helpful if your pump tracks IOB and you use a bolus calculator feature, but if you take injections, keep in mind about the insulin you’ve already given for the meal – just subtract that amount (1u in above example) from what you’d otherwise inject for the meal.

Note: if you use eating soon mode, you might want to delay the last unit or two of your meal insulin until after you see BGs rise, since sometimes you need less total insulin for the meal if you get insulin active early. (Often, we PWDs may overcompensate with more insulin than we need because it’s not timed correctly compared to the carb absorption rate.)

Example:

  • 5pm – You’re planning to eat around 5:30 or 6pm. Your BG is 120 and your correction ratio is 1:40. Setting your correction target to 80, that means you take 1u of insulin.
  • 6pm – You sit down to eat. Looking at your meal, you see 45 carbs and decide, with a carb ratio of 1:10, that you would take 4.5 units for the meal. Keeping in mind your earlier bolus of 1u, you end up taking 3.5 units for the meal. (4.5 total – 1u prebolus = 3.5 more units needed to cover the meal, see above note about considering delaying a unit or two of that bolus until you see your BGs impacted by carbs).

Result? You should have less of a spike from your carbs kicking in 15 minutes after you eat. It won’t always completely eliminate a spike, but it will provide a flattening effect. This is part of how I’m able to eat large (like 120g of gluten free pizza) meals and have flat or mostly flat BGs, and this is also one of the reasons I think using #DIYPS has dramatically improved my eAG and a1cs.

(See another post, with illustrations, about doing eating soon mode here.)

#DIYPS & #OpenAPS

Since I‘ve been using #DIYPS for over a year and also had the closed loop version running for more than two months with excellent results, I get several questions every week about how/when we’re going to make it available to other people. #DIYPS is an individual implementation that we built, and because of FDA regulations it’s not something we can give to another person to use. (Not to mention it’s not been tested for more than n=1, etc.) But, both Scott and I are passionate about moving diabetes technology forward for all, and so this week we kicked off the OpenAPS project.

#OpenAPS is our initiative to build on the #DIYPS closed loop work and eventually make this type of technology available (and faster than the market and traditional research is otherwise moving) for more people with diabetes. We aim to encourage other independent researchers to build their own closed loop implementations based on the OpenAPS reference design, and share their results and help us improve the design further. We are also working toward clinical trials that will enable more people to test and use the system during the research phase, but without having to code and build their own implementation of a closed loop artificial pancreas system. And all of this will be done in an open, transparent way so people can ask questions, monitor progress, and get involved at various stages.

The Open Artificial Pancreas System (#OpenAPS) is an open and transparent effort to make safe and effective basic Artificial Pancreas System (APS) technology widely available to more quickly improve and save as many lives as possible and reduce the burden of Type 1 diabetes.

We believe that we can make safe and effective APS technology available more quickly, to more people, rather than just waiting for current APS efforts to complete clinical trials and be FDA-approved and commercialized through traditional processes. And in the process, we believe we can engage the untapped potential of dozens or possibly hundreds of patient innovators and independent researchers and also make APS technology available to hundreds or thousands of people willing to participate as subjects in clinical trials.

At the end of the process, we hope to have produced an FDA-approved #OpenAPS reference design and reference implementation that can be used by any medical device manufacturer with minimal regulatory burden. We believe this will in turn allow manufacturers (and the academic research teams they work with) to turn more of their attention to designing and testing more advanced APS systems, and thereby accelerate the pace of innovation toward new and improved Type 1 diabetes treatments, and eventually a cure.

In the mean time, it will make basic overnight closed loop APS technology widely available to anyone with compatible medical devices, thereby reducing the burden of Type 1 diabetes on everyone who lives with the disease.

I’ll continue to post here often with data and updates from my experience & work with #DIYPS, which I’m continuing to use. But I also encourage you to bookmark OpenAPS.org if you’re interested in watching that work move forward, too – and as always, we’ll be on Twitter with #DIYPS and #OpenAPS as @DanaMLewis and @ScottLeibrand (and you can email us for #DIYPS or #OpenAPS info at Dana@OpenAPS.org and Scott@OpenAPS.org).

#DIYPS Closed Loop – One month of data & it still works great

My name is Dana Lewis and I have a closed loop artificial pancreas (that I built) that I use every night. Also see: woot!

Tweet that says "I love you #DIYPS closed loop"

We closed the loop in December (click here for more details about the first week of data & basics on how the closed loop works; or here to learn more about the basic non-closed loop #DIYPS and how far we’ve come in a single year), and I’ve been using it almost nonstop since.

When to loop – or not

I’ve only been limited in using the closed loop for overnight when I inadvertently fried the SD card in the Raspberry Pi (a known issue that’s related to the number of times you plug/unplug it from power source). We promptly ordered a second Pi as a backup, and by the time it arrived, Scott managed to revive the first one, so I now have two ready to run at any time. (Because of course you need emergency backups of your essential organs!)

When I first started looping in December, I originally intended to only use it at night, but it was hard to not loop during the day, too. The idea that I could sit at my desk and work for 3-4 hours straight, and not worry about my BGs, is so appealing!

However, it’s not really practical for me to want to use 24/7 every day…yet. This is where the versions of APS that will come to market in 3+ years will be great, because it will be all-in-one (or at least limited in excess pieces you have to carry). To wear it 24/7, I have to have enough batteries to power the raspberry pi and keep it in range, which means constantly picking it up and moving it around, carrying it in my bag from meeting to meeting, etc. I’ve decided it’s not worth it for every day, but it is awesome to have the choice when sitting through an important interview to have it running and have more security and peace of mind.

Back to using the closed loop for overnights

When we first showed our closed loop efforts off, we had about a week’s worth of data that already showed improved overnight blood glucose (eAG) and time in range, plus reduced number of #DIYPS alarms.

After a month, the results of using a DIY closed loop artificial pancreas system overnight continue to look promising:


We also recently added another screen to #DIYPS so I could access and review “what happened last night”, or what was the output of the closed loop. I usually look at the net impact per hour (how much above or below my normal basal rates I ended up getting), but also at the individual numbers of how much above and how much below I ended up getting every hour. Here’s a visualization:

Tweet illustrating how DIYPS closed loop works overnight with micro corrections to keep blood glucose in range

Using closed loop activity to spot trends and change basal rates

We also built a 7 day and 30 day view of this closed loop activity history so I can look at trends by hour overnight (versus manually compiling and visualizing them, like I did in our first closed loop data post here).

When we first looked at the data for overnight closed loop net insulin activity, I could spot a trend easily – the loop temped low from midnight to 2am, leveled out, then increased with higher temps after 5am. I can easily see why it often ends up temping low after I go to sleep, because I often do a half unit or so if I’m riding “high” (130 – ha!), and feel safe doing so with the loop to catch any more drop than I desire. (This pattern was consistent at 7, 14, and 30 days, although the average net insulin amount varied slightly.)

However, when I first started looping overnight in December, I had rises around 2am that the loop had to handle by providing higher temporary basal rates than my scheduled basals. I upped my midnight basal rate as a result, but it looks like I upped it too far since the closed loop often temps low now during that time frame. My next step is to probably change my midnight basal to be lower by 0.1u, and see if that reduces the amount of low temps needed from 12-2am, without causing spikes. My theory is that this alone may also reduce the number of higher temps from 2-4am, because those may be a result of the lower temps earlier on. Thus, I am starting overnight basal tweaking with this one change and seeing how much of a ripple effect it has for the rest of the night as well.

In pursuit of…what? Or: what are we optimizing for?

My BG averages overnight (and during the day) are excellent (well within the “normal” range <120) and time in range is continuing to grow (85% & counting) with using the closed loop #DIYPS. So, beyond minute basal tweaking, what’s next?

At this point, I think the next goal is encouraging other independent researchers to get up and running on their own closed loop overnight APS.  Do we really need to wait another 3+ years for APS technology to come to market?  What exactly would #WeAreNotWaiting look like if we applied it to closed loop Artificial Pancreas technology research? What do you think?

Why #DIYPS N=1 data is significant (and #DIYPS is a year old!)

As I’ve said many times, last year we set out to create a louder CGM alarm system. By adding “snoozes” so I didn’t drive my co-investigator crazy, we realized I might as well enter what I was doing, and be precise about it (aided by some quick bolus and quick carb buttons that made data-entry not the chore that it sounds). Thus, we had the data and the brains to realize that this made for some great predictions; much better than what you usually see in diabetes tools because they rely only on your insulin sensitivity factor (ISF) and correction rate, but don’t take into account carbs on board and their impact over time, etc.

(If you’re new to #DIYPS, read about the beginnings of it here. For more on this idea of carbs on board and the carbohydrate absorption rate and how significant it is for people with diabetes, read about that here.)

After I had spent 100 days using #DIYPS, Scott and I stopped to look and see what the impact was. For the long version, read this post about the results and the direct comparison to the bionic pancreas trial data that was available then. The short version: #DIYPS reduced my eAG and a1c significantly, reduced lows, reduced highs – aka my time in range was improved from 50% to regularly 80+%.

I have asked myself (and others have asked), are these results sustainable? Are these improved outcomes truly because of #DIYPS? It’s definitely worth noting I never changed what or how I ate. (I ate 120 grams of pizza (for science! ;)) several times to test the system, but I didn’t eat less or any healthier or otherwise change my diet.)
I can’t attribute these outcomes directly to #DIYPS alone, but I do believe they’re highly correlated. It’s hard to separate other contributing factors like the fact that I have more boluses per day using #DIYPS (which other studies have shown decreases a1c); or the fact that I spend less time high/low because with #DIYPS I actually can wake up at night and take action before I’m high or low.
So, it’d be hard to study specific factors and say “it’s all #DIYPS”. But, I’m pretty sure it’s mostly #DIYPS. Regardless, here’s the updated data about the sustainability of the results I’ve seen with #DIYPS over the past year:
Why this is significant
#DIYPS is currently n=1 (meaning one person is the study’s subject). But what is significant is that I have year’s worth of data and actual lab-tested a1cs that shows the outcome of this type of artificial pancreas work. And it (to date, but coming soon  / OMG this week!) hasn’t even been closed loop – I’m still the “human in the loop” making decisions and pushing buttons on my pump.Compared to the bionic pancreas and other artificial pancreas study trials where they have a few days and a few more (usually n=20 or so) subjects; they can look at the decrease in lows and highs and improved eAG…but they can only project what the a1c improvement is going to be.We’ve shown the improvements in lab-tested a1cs – see my graph above?

It all adds up
Scott and I are not the only ones working on a closed loop. The community of developers connected to the Nightscout community has nearly two hands full of people who are working independently on device interoperability to close the loop by freeing our data from devices and enabling us to work with our own algorithms regardless of which hardware device we use to support our diabetes management.When all of these n=1 studies add up, it matters. At some point in the near future, after we’ve closed the loop with #DIYPS (ah! this week! :)) and others have as well, we may have more n=1 hours on closed loop artificial pancreas systems than the (traditional) “researchers”. Scott and I are hoping that we can not only show the world how open source innovation and new regulatory paradigms can deliver safe and effective results for people living with T1D faster than traditional medical device development and traditional regulation; but that we can also change how all successful medical device companies approach interoperability, and how traditional medical researchers do research – possibly in partnership with patient researchers like us.