Whilst I was more or less forced to swap off Fiasp a week or so ago, I thought it might be interesting to take a look at the various data and graphical tools that I have available to me from Nightscout and OpenAPS to see if there is anything in there that helps to explain what was going on, or gives an opportunity for early warning.
If we start with the graphs that we get from NightScout, you can see in both the distribution and percentile plot that there’s a clear issue with Fiasp when looking at the data retrospectively.
Whilst these two periods have very similar Hba1C values, the difference in standard deviation is marked, and the reason why can be seen in the Percentile graphic.
What’s interesting is that different weeks have very different distributions. The following two are from 14th-20th August and the subsequent week:
The percentile graphs show a similar story:
It very much suggests, from this n=1 set of data, that Fiasp can be highly variable in its effects when you’re doing mostly the same things.
I thought, as a secondary point of view, I’d take a look at the IC ratio throughout August and September to see if that provided any indication of what was going on.
Using the data captured from OpenAPS, I’ve adjusted all the daily basal rates to an average, and either added any additional over the average to the bolus insulin or subtracted any under the average from the bolus insulin to provide an estimate of basal. Whilst I realise this could be changing, because of the way that SMB works, using part of the bolus to cover basal, this felt like a reasonable estimate to try and determine how much insulin was only used for registered carbohydrate. I have also removed days from the data set where I was relying on the Unannounced Meal function (UAM) as I don’t have decent carbohydrate records for those days.
The graph of variance in IC ratio is below:
Nominally, this value is set at 1u:9g of carbs on my pump, but it’s very clear here that it was rarely at that level during this period with Fiasp, and that most of the time it was more like 1u:8g, or worse. Fortunately, I had OpenAPS handling a lot of the downside there, but it goes to show that with such a daily variation, living with insulin isn’t that simple.
What’s also interesting is that if you look at the dates from the earlier distribution and percentile graphs and track those on to the IC ratio variance, you see that the periods with what appears to be the greatest instability are those where the IC ratio drops and therefore more insulin is needed to handle the same amount of carbs. Now this could have a regular physiological cause, such as stress or something else, but that it seems to happen with a reasonable amount of variability, and no apparent reason makes it difficult to spot and handle.
What I find notable is the instant change on the 17th September with the move to Novorapid, where the IC ratio jumps up to 1u:9g and appears to have stayed there. Indeed, when looking at average basal insulin used, over this period, it has reduced by some 3u per day (approx one sixth).
Now given the IC ratio changing like this, it’s fair to assume that Insulin Sensitivity is also moving around a fair amount, and that makes many daily activities much harder, such as delivering accurate correction doses.
The final point on this graph is the period from about the 12th to the 16th September, which is the period in which I dropped from nearly 1:8 to 1:4. This was when I made the decision to revert back to Novorapid. I felt that I was constantly chasing and it wasn’t working very well.
Now it’s possible that these are the symptoms of a batch of insulin that has gone off and shouldn’t really be being used. I’ll find out soon enough when I get a new one, but if I take a quick look at the days with the largest bolus doses, they don’t align with insulin batches, indeed you can see that while the last few days show very large bolus requirements, there are plenty interspersed through the 8 weeks of data:
The top ten doesn’t include the couple of occasions where I (or a dog) had managed to remove my cannula.
It all goes to show a complex relationship between a number of factors. It also doesn’t explain what happened in the slightest.
Are there any conclusions that can be drawn from this? Well I think the one thing I’d take away is that if you feel like you are needing more and more insulin for the same amount of carbs, and it never stabilises or reverts to a normal position, then it’s perhaps time to make a change. It’s not always easy to see that though.
This period also shows how effective OpenAPS is. While it was clearly not that straightforward, the system managed to, for the most part, keep me on track. With the variability that is shown there, that’s significant.
Another I’d draw is that the two week trial period for CSII in the EMA Submission paper probably wasn’t enough. Most people seem to be seeing issues beyond that point. It also raises the question of how long Lilly will test their “faster Humalog” which uses an excipient to locally dilate blood vessels in order to accelerate absorption. That sounds like something that I wouldn’t want a build up of to start doing unusual things.
Finally, I’m not entirely convinced that the increase in reported occlusions that proved not to be in the EMA submission are not related to the same thing that I’ve seen here, and that others have reported. Quite what that is though, is a very different question, and not one I have an answer to yet.
Suffice to say though, that whilst my first few months on Fiasp felt like they held loads of promise, the latter few weeks felt like a bit of a roller coaster, and the change when reverting back to Novorapid was like Night and Day. That’s why I’m keen to find out whether it was a batch issue, or whether I have some physiological aversion to it. I hope it’s the former, as the removal of a need to pre-bolus has been incredible. Unfortunately, I’d put a bet on it being the latter, which would be a bit of a downer and put an end to what seemed like a great new insulin.
For me at least, it seems that for now, the idea of a bolus free artificial pancreas is still a little way off…