One of the items that stood out about ATTD was the amount of representation about DIY systems at this year’s conference. The number of posters talking about DIY APS systems hit the teens, and one poster was selected for an Oral presentation.
In addition, an initial set of Jaeb study data was presented. For those not in the know, the Jaeb study aim is to collect information from at least 300 – 1,250 adults and children with T1D using Loop in the United States. This information will be used to learn more about how well Loop works, what problems users have, how often severe hypoglycemia and diabetic ketoacidosis occur, and how well the system controls blood sugar levels.
Using the Jaeb study data and one of the posters, I’m going to look at Automated Insulin Delivery systems, and why choice matters.
Jaeb Study Interim Results
The results so far were published on Friday 21st February and were very encouraging. New users (people in the study who had never used this closed-loop system before) showed:
An A1C reduction from 6.8% to 6.5% after three months and to 6.4% after six months – and if you think that is not very much, think again!
An increase in time in range from 68% to 73%, which is more than one hour per day spent in-range! These time in range benefits occurred in the first month of closed-loop use and were constant throughout the rest of the study.
Benefits in A1C and time in range across all age groups.
Improvements in user-reported outcomes, including measures of diabetes management distress, sleep quality, and fear of hypoglycemia. Not too many details were shared on this part and we look so forward to learning more!
It’s worth paying attention to some of the baseline data associated with this study, as it also showed that those who had participated in the study were classified as having “well-managed” diabetes at the start of the study, with a relatively low A1C (the 6.8% baseline was substantially lower than any of the Commercial studies and a whole 1.4% below the CamAPS-FX trial). Time in range goals were also close to being met pre-looping and the participants tended to come from high education levels and socioeconomic status backgrounds.
But even with the above statement, users got better outcomes, and significantly, felt better using Loop, which is a key factor in all of this.
The study runs until the end of March 2020, so we won’t get full results for a little while yet.
Then we have the “Pig Experiment“…
What’s the “Pig Experiment“?
It’s a study run by Rayhan Lal attaching Loop and AndroidAPS to pigs.
As the poster reports, there are ethical concerns about attaching unregulated devices to humans, which is why they selected pigs. It’s worth noting that there are plenty of differences between pigs and humans, not least of which is that insulin acts more quickly in pigs.
The pigs with Diabetes (#PWD – who’d have guessed) had their starting settings assessed clinically to provide ISF, CR and basal rates, and then the systems were attached and allowed to run. During the trial the pigs were fed a high carbohydrate meal in the morning and then regular pig chow for lunch and dinner.
The study hypothesis was that AndroidAPS would handle the meals better than Loop due to the UnAnnonunced Meals (UAM) and Super Micro Bolus (SMB) features.
The results of this trial supported the hypothesis, remarkably well it turns out.
When the pigs used AndroidAPS, they got 63.7% Time in Range (TIR) over 23 days while when they used Loop they got 40.5% over 18 days. I’d hazard a guess that when the pigs used AndroidAPS they were probably the happiest.
Neither is a long period of time, and the participant group is very small, but the key thing in all of this is that pigs are unable to use phones and therefore unable to intervene, so we are seeing the algorithms working to the best of their abilities with no human intervention. And many would argue that Loop and AndroidAPS are both Hybrid Closed Loops rather than fully closed loops, so the test is irrelevant.
As the average glucose images show, the post-prandial period was when the effects of the different algorithm components appeared to be most critical.
What has one got to do with the other?
And there’s the crux of this. The two items appear unrelated, and indeed, experimentally, are. But the outcomes are worth investigating.
The Jaeb users show a TIR increase from 68% to 73% when new users moved on to Loop. The Pig study showed a TIR of 40.5% with no baseline for the pigs on Loop.
While the Loop users saw an increase in TIR they also reported better user outcomes. Loop has clearly made life easier, but what are the important things that they get from Loop that they feel has helped and how much of what they used to do before have they been able to drop? We should find out more on this in the full results. Have any of their behaviours not changed, but do they care?
And more importantly, why did they choose Loop? Was it because they had an iPhone, or because they used Omnipod? Was it because it had a better interaction model? Because they knew they had to interact with the system to get best results so chose the one they preferred?
On the other hand, the pigs don’t really understand what a phone is and they almost certainly don’t care about user interaction. For them, what probably matters most is how they feel. In that context, either system is better than none, but they’re going to have to use it as a fully closed loop so whichever does that best is likely to be preferable. Given that context, they’d probably take what they could, but if they could choose, I’d expect them to choose AndroidAPS as it probably made the feel better.
Ultimately, aligned with the point that was made (many times) at ATTD2020, one closed loop doesn’t fit all. What this seemingly disparate data shows is how important it is to understand what is really important to an individual in picking the system they want to use. What’s the context of the user?
These cases show that factors such as what phone you might have, what type of pump you might prefer to use and what it’s like to interact with also play a part in gaining better management. The user experience of Tandem pumps has also supposedly backed this up. The question becomes, what is “good enough”?
For me, the following slide from ATTD 2020 sums it up nicely:
And as much of a surprise to some as it may be, having a more effective algorithm may not be the key driver of that decision…