When making changes to our diabetes management we sometimes see changes in our total insulin consumption. And it’s easy to jump to conclusions regarding cause and effect. Whether it’s a new insulin, a different diet, a new adjunctive treatment, changes to exercise and body composition, or whatever. It’s easy for example to look at the numbers from one week to the next and try to infer relationships. But it’s never quite that simple.
My AID systems have been keeping fairly detailed records for years, and it can be educational to occasionally review it in search of patterns. People sometimes ask generic questions like “What’s your TDD?” (Total Daily Dose). As though that precisely defines something about us. Of course the answer is to some extent: It Depends!
I believe that looking at large data sets visually and not just as a list of numbers can help our brains comprehend them. Here you can see three and half years of daily TDD records:
First thing we see is that it jumps around a lot. Even when we try to smooth it out with a rolling 10-day average, it still oscillates.
Today I think of my TDD as being in the 30-40 U range (currently at the 40 U end). But even that’s just a “ballpark” range and I know it’s going to change.
Now, a few of the individual points are flawed. For example I remember that on that “74.41 U” day in January 2023 a failed cannula meant that most of that insulin didn’t actually go into me. But for me such events are rare, and I haven’t filtered them out of this graph. Also, when travelling across timezones the length of “a day” can vary, which if we’re not careful can skew the data.
I do have lots of other data recorded, and I do see vague correlations with things like weight (for example I weighed more at the end of 2022 and there’s a rough correlation with both lower ISF and higher TDD around then). But that doesn’t explain some of the other rises and falls. There are obviously many factors interacting.
Incidentally in the AID system I use, the instantaneous ISF is currently affected by the current BG levels, but not the recorded TDD. That works well, and trying to include rolling TDD data in the calculations introduced too much variation in the ISF.
I put most of the day-to-day variation down to diet. Basically if I eat more I end up using more insulin. And I don’t live a life constrained by carbohydrate targets (I started using a fully-autonomous AID system in 2021). Things like fasting for colonoscopies and days of big bike rides do result in lower insulin use, but those aren’t every day.
The peak in late 2024 does correlate with an increase in my HbA1c as well as other measures such as decreases in my Time In <various> Ranges. That didn’t surprise me though, as I had been participating in several clinical studies over that time (including studies into new CGM sensors where they needed me to run my BG well outside my normal range at times). I’m not suggesting that there are direct associations between these measures other than all being signs that it was “a fairly experimental phase”…
Is this useful data?
I’m sure there are very few people who actually look at data sets this large. Even if they have them. Is it silly?
First thing to note is that I don’t spend all my time recording data. Or looking at it. The software running in the background does all the gathering, and I only look at it occasionally. And I do find parts of it useful.
There have been changes in medication and management along the way that seem to have contributed. But it is sobering to see how much the data can “bounce back” after a change and invalidate early assumptions about impact. So it can take some time to build confidence in hypotheses.
Clinical studies into such things do try to have large cohorts to average out all the individual variations between people. But often they still end up sampling fixed end-points in time for each person. Make the cohorts large enough and it can increase the statistical “power” of the study, but even so I do feel there can be insights found from also observing individual data sets. So I’m not going to throw my data away any time soon!