One of the things that I’ve previously noted is that I’ve seen a fair amount of variability in using Fiasp, both in insulin sensitivity and in mealtime sensitivity, best described by Carb Ratio.
What I’d felt particularly was that during the week, I was needing substantially more insulin for food than at the weekends, but I hadn’t done anything to really check this.
As I’ve mentioned previously, one of the huge benefits of having the various DIY Loop set-ups is that you have a great deal of instrumentation easily available and this allows you to understand quite a lot about what’s going on with your body/diabetes/insert what you want to know about here.
So coming back to the earlier point, I have two mechanisms for understanding variability in Carb Ratio. I can either take it from the daily Autotune run, or extract the data from NightScout and manipulate it. The latter should broadly match with the former.
If we take a month of data then, what do we see? The below:
What’s rather disconcerting about this is that it is very up and down, with some days showing very low ratios and others very high. There is, however, almost a pattern to be seen, with weekends tending to look like they have more carbs per unit of insulin compared to the weekdays. In the graph above, the Orange line is the average.
If we follow through on the above mentioned hypothesis, that weekends require less insulin per gram of carbs than weekdays, it would make sense to do a clustered view of the data using each weekday as a node, and potentially weekdays and weekends as additional nodes.
This gives us the below:
Again, the orange line is the average carb ratio. This identifies a very clear pattern, which is that Saturday and Sunday appear to require less insulin for food than most weekdays, with Monday and Tuesday need ing less than the second half of the week.
Comparing the average weekend day to the average weekday day, we see that the difference is roughly 12.2g/iu to 9.5g/iu. Now you may ask whether there is a correlation between the amount of carbs that I eat per day of the week and the variation in carb ratio for that day, in that on those days where more carbs are eaten, the g per insulin unit becomes lower. What is worth noting is that the standard deviation of the data on a daily basis can be quite high, so it’s not a perfect analysis.
The relationship isn’t particularly strong in that respect and whilst there is a correlation it is a negative correlation, and not a very strong one, with a value of -0.16. So it would appear that on days where I require less insulin per gram of carbs, I eat more carbs, which makes sense, as I am likely to be avoiding hypos. What I also note is that carb consumption has a standard deviation of about 12%.
What can we take from this? Well, it’s clearly not all about the Fiasp. Whilst I may have noticed more variability on Fiasp, there is clearly something additional going on.
Based on the rather striking visual above, what I take from this is that a combination of tiredness, stress and sedentariness play a part in how my body interacts with insulin. Whilst I may walk 10,000 steps most days, weekdays sees this generally split into two periods as I am on my way into work and on my way home. I am then sedentary throughout the day, whereas at the weekends, the steps are more evenly distributed throughout the day, and I am doing more activities, instead of sitting at a desk.
Secondly, as the week progresses, it’s fair to say that job stress is definitely higher than when not working, until late on Friday, when I have a weekend to recover. So stress hormones will be playing a part.
Finally, weekday sleeping patterns are constant, but rarely exceed five hours, whereas weekend evenings (Fri-Sun) generally result in six to eight hours of sleep. it appears this benefit is carried over into the early part of the week.
Conclusion
For me, in my n=1 sample, it would appear that during the week, I’d benefit from more sleep and managing stress slightly better. What it also shows is that while we may attribute variation to something like the insulin we use, there are often other factors at play, and we do need to consider those when looking at the overall picture.
Finally, it adds to the questions we need to ask about how we go about calculating the various numbers that we use to live by. Clearly, from my perspective, I’ve been seeing a lot of variation, and it’s driven by a number of factors, but it also suggests that those days where we just can’t seem to avoid hypos are often driven by physiological factors. Remember that many pumpers use a different basal pattern at weekends? Well maybe they should also use a different ISF and Carb Ratio as well. For some, it may even suggest that daily ones make sense!
The more instrumentation we get of the data that we can produce the more that T1D strikes me as being similar to a modern fighter jet. Inherently unstable, the pilot (us) is asked to fly it with the aid of multiple tools. The better the tools and understanding of how to use them are, the more stable the flight…
The comparison with an aircraft is interesting and relevant because both are non-linear multi variable control systems. The assumptions around ratios when controlling blood glucose are somewhat flawed because there is an assumption that their effect is linear when it isn’t. However, linear approximations to non-linear systems can work when the numbers are small.
In a control system like an aircraft inputs are integrated with respect to time. In other words, the size of the input affects the rate of change of the output, not the output itself. I suggest that the same is true of diabetes, you need to think of the plasma insulin level (concentration of insulin in blood) affecting the rate at which glucose is removed from the blood. When the numbers are small the size of bolus may work as a proxy for plasma insulin level, but only approximately.
I suggest that you need to estimate plasma insulin by integrating insulin inputs with respect to time taking into account the absorbtion profile as given by the suppliers.
Thanks for your comments Richard. What you suggest for estimating plasma insulin is basically what’s being done in the DIY loops to provide a reasonable estimate of insulin on board.
I also think that the response to insulin is non linear. For example, a very high blood sugar will not respond to insulin in the way you might expect in terms of correction bolus. My experience is that you need to raise the plasma insulin above a certain level to get it to shift at all, then as the BG comes down it accelerates, until maybe the drop has to be checked by some sugar intake. I’m thinking of an equilibrium reaction in chemistry where things start to move one way or the other when the concentration of one side is sufficient to match the concentration of the other side of the equation.
I think most people with T1 would agree with you, however, instrumenting and modelling that is a tough ask given the sensing technology and insulin action times we currently have. Some of the diy systems do attempt to do this to some extent.
Reminds me of Scott Hanselman’s “Diabetes Explanation- the airplane analogy”
https://www.hanselman.com/blog/ScottsDiabetesExplanationTheAirplaneAnalogy.aspx
Great article. I’m extremely curious to see how the same charts would look on Novorapid as opposed to FIASP and how much is to do with the insulin as opposed to other factors. For myself, I have switched from FIASP back to Novorapid in my pump as not having the benefit of a looped system I was finding the variability very frustrating. How much of that is to do with the insulin and how much is to do with other factors is really difficult to pin point. I’ve only been on a pump for a few months and have found it much more difficult to keep things in range than when I was on MDI and for me not knowing if that is the use of FIASP, the pump regime or something else I decided to remove FIASP from the equation and things are seeming to be a bit more predictable at least. When using FIASP in the pump I found I still had to pre-bolus like Novorapid and when it worked it did seem to do a great job of preventing post meal highs but it became very random as to when it seemed to work and when it didn’t and I had several evenings of feeling very awful with high levels which I struggled to bring down. Maybe I’ll go back to FIASP at some point but for now I understand it is not approved for pump use in the US and Canada and personally I’d like for a bit more understanding of the implications of the niacin additive when using FIASP to cover all insulin requirements rather than just bolus requirements (I’d be using FIASP succesfully in pens for about five months with good results pre-pump).
My suspicion is that with Fiasp, we are seeing a more extreme reaction within a pattern that probably normally takes place. It’s fairly difficult to prove, although we have some ideas!