A month after this post, I’ve now completed a two week run using a single basal rate with circadian like ISF and CR adapted into it.
The original plan was to run with a mix of Fiasp and Lyumjev, whilst announcing and partially bolusing for meals. It quickly became apparent during the first two week warm up process that that’s not really how I use the system, and I tend to bolus and then leave UnAnnounced Meals (UAM) to handle the rest of the meal, so the testing was adapted to reflect this reality. I had also switched over to an adapted version of the code during the first two weeks that would better manage UAM. Additionally, I decided to stick with Lyumjev only as the insulin I would use. Target glucose throughout was 6.0 mmol/l. Average carb consumption over this period ranges from ~150g to 250g per day.
The basal rate that was used was calculated by taking the previous seven days total daily dose at the start of the process, multiplying by 0.4 (given that’s the estimate of how much of my dosage is from basal) and then dividing by 24. The resultant value was approximately 30% higher than my overnight basal rates, so instead of using it, the overnight basal was used, for safety reasons in case the pump and system lost connectivity overnight.
To calculate the CR and ISF, the principles in this paper were used: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3879757/. This resulted in a highly varied CR and ISF throughout the day, as can be seen in the next graph:
I’ve included the old basal rate as the yellow dotted line to provide a comparison. That was run with a single ISF and CR. As can be seen, the variation in ISF and CR is fairly significant over the course of 24 hours.
The glucose distribution (TIR range 3.9-10 mmol/l) is shown below.
The TIR of 87.8% isn’t widely different from that I’ve seen previously while using a DIYAPS system. If we look at the AGP, we see that whilst overnight was fairly tight, the daytime operates in quite a wide range:
We should bear in mind that this data has a couple of issues, in that there is a Dexcom replacement that resulted in some wildly incorrect readings over a period of about 6 hours that has skewed the high data.
If we want to understand this better, the TIR for 3.9-8mmol/l is shown below.
As we can see here, the TIR above range doubles with the tighter range.
With the variable ISF and CR, the system seemed well able to handle the single basal rate in normal circumstances, as the adjustment in ISF more than evened out the lack of variation in basal rate.
The biggest limitation with a single ISF rate is that when you rely on UAM for handling a proportion of meals, the time periods where you have the lowest ISF numbers, i.e. when you’re likely to need greater basal rates are more limited in the amount of insulin that can be supplied in an automated fashion, due to the limitations of the limits on multiplying basal rates. As a result, there were times where I felt that I required more manual input than perhaps I did previously.
On the other hand, I really don’t think this would have worked all that well with a single basal rate, single ISF and single CR. I know there are times of the day where I am more resistant than others, and I think that variability is required throughout the day.
What I do note is that my old model with one ISF and CR and multiple basal rates probably needed some level of review, given the vast difference on basal rates in use.
Did this set-up work for me? I think the answer here is that it did, but it limits some of the automated insulin delivery tricks that are available in AAPS. Whether it really is that different in terms of outcomes, well, we’ll find out in the next month or so as I test it all with a Nauck Circadian profile…