As readers of this page will be aware, I’ve been following Ultra Rapid Lispro for a little while, as Lilly has published various data at a number of conferences. Following its EMA approval earlier this year, and launch in the UK in July (all be it a very quiet, nay, almost silent one), I’ve managed to get my hands on some and now have a chance to see how well it performs with OpenAPS.
At ATTD in Madrid, we learned that Fiasp in commercial AID systems wasn’t any more successful at maintaining time in range than the traditional, and off patent, “rapid” acting insulins, which we surmised was due to the insulin models the commercial systems use.
We learned from Profil that Lyumjev appears to peak between 35 and 45 minutes:
And that it also appears to have a considerably shorter tail than nearly all the other fast insulins out there (very interesting from an AID point of view):
Finally, in their meal tests, they found that it was as effective as the pancreas in eliminating post-prandial excursions for two hours after the meal, however beyond that, it wasn’t quite so good:
This all looks very exciting.
The clinical trial that was run to investigate use with a pump threw up some concerns:
We can see that Lyumjev (original code name LY900014) showed 3x the number of adverse events that regular Lispro did. Of this, infusion site pain was more than four times more likely while a reaction at the site was a little over twice as likely. In spite of these issues, only one participant dropped out of the study, which suggests to me that for the users, the benefits outweighed the issues.
To put this in context, in the Fiasp pump trial, there were twice as many pump site reactions as regular Insulin Aspart. In other words, it appears that Lyumjev users are more likely to suffer reactions or pain at infusion sites than Fiasp users. However, the number of issues that resulted in the users changing the sites early were no worse than Humalog (if anything, the data suggests that Lyumjev required fewer irregular set changes):
With Fiasp, there was a 23% increase in unexpected changes to infusion sets.
This is encouraging in relation to the experiences that we have seen with Fiasp and unexpected hyperglycaemia that doesn’t make sense.
The upshot of the above summary is that in the patient leaflet, site reactions or pain are listed as common side effects, as they are with Fiasp.
Using Lyumjev with OpenAPS
Given all of the above, how should Lyumjev be used with OpenAPS? From the Profil clamp testing, it suggests that we need to run a custom insulin curve, which was a capability that was added when the exponential curves were added to all the systems. With the clamp testing data presented, my initial settings for the curve are:
|Duration of Insulin action||6 hours|
|Peak time||45 minutes|
Now, you may be asking “Why use six hours DIA, when the graph appears to show five?”. This allows the algorithm to think there is insulin on board when there may or may not be, and is a safety factor. As I observe behaviour, I may adjust it down. We can also see from the clamp graph that the profile for Lyumjev is slightly different from that of the other insulins, with the flattish peak that lasts from around 1 hour to 3 hours. Time will tell whether this is required to be modelled into the systems.
Secondly, for the run in period, I do not be adjust any profile settings before use. Instead I use a temporary target of 7.5mmol/l to leave space for landing should the insulin turn out to be more effective than the Fiasp/Humalog mix that I have been using. As I use OpenAPS with the option that a raised target tells the algorithm that I am more insulin sensitive, it also means that the system will run with a less aggressive ISF and therefore will be more conservative in dosing insulin.
With all of this in place, at the end of the day, it runs Autotune and we can see what gets adjusted. In this case, the starting point is:
After one day of Lyumjev, the adjustments that have been made by Autotune can be seen below. Looking at total numbers, the answer is by not very much, although the I:C ratio has dropped, which feels as though that’s in line with my first day of using it. Digging into the logs from Autotune, it’s possible to see that the calculated full change in CR is from 16.4 to 11.4 (Autotune attenuates the adjustment for various safety reasons), which is a pretty significant difference. Likewise, it’s also picked up a pretty big change in sensitivity.
But. How fast is it?
I’ve not figured that out yet…. As the picture below shows, a 2u bolus taken with glucose levels at around 6mmol/l doesn’t result in a sudden drop within 30 mins, or even 50. IOB at point of taking the insulin was approximately 1u, which should be cover for the instant 0u TBR that kicks in from OpenAPS with the bolus. Measuring the speed turns out to be harder than expected.
Having used Fiasp and recalling the immediate, obvious impact that I saw with it, it doesn’t feel like I’m getting the same thing with the Lyumjev. It doesn’t feel like it’s particularly early, in fact, it really feels like it’s a bit late.
There could be reasons for this that aren’t the insulin + additives, of course, and I probably need to try a different vial to see if the one I’m using isn’t quite right.
Ultimately, 24 hours isn’t really long enough to adjust me or the system, so let’s see how it plays out over the first week or so, and whether Lyumjev makes a major difference.