I hope the following discussion is also interesting for pumpers. It shows a potential new feature and I would like to hear your feedback.
The Insulin On Board (IOB) is a guestimation how many units of insulin are still active from the last shot. Foremost it is a reminder that additional physical activity might amplify these remaining units (risk of lows). In addition to that it helps to prevent the overcorrection of a higher blood glucose.
With physical activity I see the need for an additional reminder: Sensitivity On Board (SOB). After sports the sensitivity to insulin is increased for several hours. The muscles will restore their glycogen stores and this will for sure lower the blood glucose. The timespan depends on the level of activity and individual reactions. For simplification I will adopt the workings of our Glucosurfer project but I think it can be applied to a pump too. In our diary we have four levels of activity: normal, moderate, high and maximum. Let us assume we can define the SOB timespan in minutes for every level.
Now the system would work like this:
You would work out for 1 hour and this will be documented in the diary / pump: high activity starting at time x. Associated with this high level you have defined that high activity leads to an increased sensitivity for 300 minutes (=5 hours). So for the next x+5 hours the SOB warning will be shown. It will remind you that you are now acting under the influence of physical activity that happened hours before. I am not sure if it is wise to use the information to directly reduce the I:C. Maybe the character of a warning would be more appropriate. What do you think?
That's a good idea although I've noticed some variability? My winter "regimen" was alternating lifting+ elliptical w/ running although it was highly weather dependent. I've noticed that after lifting hard, I get a bump in my BG, which I attribute to adrenaline but that afterwards, the insulin seems really zingy? Sometimes I'll eat more and sometimes I turn the basal down but it can be a bit of a bother.
Ginger Vieira wrote in her book "Your diabetes science experiment" that powerlifting has two opposite effects for her. First it does increase the blood glucose because extra glycogen is released. This can be accompanied by the release of cortisol which will blunt the insulin sensitivity. But after this phase the sensitivity to insulin will be much higher for several hours. This might be in line with your experience.
But back to the SOB. Would it be sufficient for you when your pump would show something like this in the display:
⚠ SOB 68
Interpretation: Warning - remaining Sensitivity On Board for the next 68 minutes!
Well, I'm not sure I would explain things that way. I believe exercise increases the number of GLUT4 transporters which are the key mechanism for uptake of glucose. And for me the cortisol and adrenaline during lifting cause my liver to dump glucose. After lifting, I am certainly more insulin sensitive.
And for me there are actually two areas of concern, the time after exercise in which you not only have increased insulin sensitivity, but you also reload depleted glycogen stores. This first phase is about 2 hours for me and I actually need to eat after exercise.
And then the second, which is an increased insulin sensitivity over the next 24-48 hours. I do think everyone will be different and you will respond to different types of exercise in different ways.
In can not tell if it is the cortisol to reduce insulin sensitivity or the adrenalin to secrete more glucose but the outcome can be some sort of spike afterwards. She would bolus normally for the first meal directly after lifting. Then she would reduce her dosage for the following hours. I do not see this kind of reaction but perhaps it depends on the intensity. So in my case I would always reduce my dosage.
Do you think the SOB warning can be implemented in a way that 70% of the real world scenarios can be covered? For me this is the magic hurdle. To cover the last 30% will be over the top. I think the pump developers will apply a similar thinking. Let us assume the duration of the SOB can be set for every level of activity:
moderate: 180 minutes (3 hours)
high: 300 minutes (5 hours)
max: 480 minutes (7 hours)
Of course you can extend that if necessary. But will it be useful to be warned for more than 24 hours? Or would it be better two have two phases that lead to different types of warnings.
moderate: 60 minutes (alert) then 120 minutes (warning)
high: 120 minutes (alert) then 180 minutes (warning)
max: 120 minutes (alert) then 360 minutes (warning)
The two phase design could also be used to automatically reduce the dosage calculated by the bolus wizard. In the working of the Glucosurfer this would mean that the alert level would automatically set the high level of activity to calculate the dosage accordingly. Later in the warning level the activity would be set to moderate to reduce the dosage only slightly. Does this sound like a compromise to follow?
In my mind I have played with the idea that features proposed here might catch on to the manufacturers. But your remark is totally right. The main reason for them being conservative about these innovations is that they are developing under the constraints of medical devices. We should not forget that when we critizise them for being kind of slow in adopting innovations. They have to fully understand and medically justify the inner workings of their devices. This is similar to the constraints of the aviation industry.
For the Glucosurfer we are in a more comfortable situation. Although the lobbying of the industry has already begun to open fire on these projects. Obviously the industry wants to get rid of these Apps by forcing them under the medical device umbrella. As a community we should keep an open eye on these legal developments. Unwatched this might hurt our interests in the future.
For the implementation we try to keep it as simple as possible. We already have a percentual reduction of the total insulin dosage in dependecy of the level of activity. Thus with Sensitivity On Board we would let our users define their own percentual reduction for the phases alert and warning (see above).
You further refined that to let the user define the aerobicness of the excercise. That is interesting but adds additional complexity. I am tempted to put that more on the 30% side of real world scenarios we will not cover. Thanks again for your thoughts. This is why I am asking you all. Keep it coming...
In my view, a complete IOB model should account for insulin both from carb and from correction boluses. And then the IOB should be accounted for in subsequent carb and correction boluses. Many pumps still only do part of the job, for instance the Omnipod apparently only keeps track of IOB from correction boluses although it does account for the IOB in carb and correction boluses.
The trouble with keeping track of IOB from carb boluses is that in order to do it properly, you have to introduce the concept of Carbs On Board (COB). If you bolus and eat a meal and then an hour later test, at that time you will both still have IOB and you will likely have COB. So how do you figure out whether a high blood sugar reflects a need to correct or simply that your insulin has yet to take action to cover the meal. That is the quandry and why most insulin pumps have not yet embraced this concept. But it is exactly this problem which drives people "nuts." You test after a meal and you are high and the pump wizard is no help in figuring out whether you should or should not correct.
Perhaps you could think about such a model for Glucosurfer.
The problem is that the IOB itself is just a generalized model of the onflooding and degradation of the insulin. It is just individualized by the duration of the insulin activity that is all.
Now imagine we are measuring 200 mg/dl after 60 minutes. The IOB will tell us that there is still 70% of the injected insulin on board. If you calculate the correction from 200 to target value it might be 3 units of correction. But 70% of the previous injection are still active and most likely this is more than the 3 units of correction. So by subtracting the IOB from the correction we will have no correction at all. The point is that this might not even be wrong (we actually do this calculation in the Glucosurfer to prevent insulin stacking). Yes, you have a high and unwanted spike and the insulin seems to come too late. Still this might level out after 4 hours - it depends on the individual unfolding of the insulin action. The spike is more a sign of a bad orchestration of injection and digestion. With a longer waiting time between injection and eating it might have been prevented. Of course it is also possible that the spike is the result of a misjudgement of the carbohydrates. This case is in need of a correction shot - but how many units?
Now imagine you see the spike not at the 60 minutes mark but at the 120 minute mark. This is much better for our calculation because now more than 70% of the IOB insulin is already gone. This means by subtracting the remaining IOB from the correction of 3 units there will be some residual insulin like 0.5 or 1 unit to inject. This small correction will most likely work as expected without causing a low.
Just another example:
Scenario 1: BG of 250 one hour after the meal. The IOB is 6 because I had a high carb meal that needed 8.5 units. The proposed correction for 250 will be 3.5 but minus the IOB of 6 it will be 0. Here the situation can not be solved easily. There is too much Insulin On Board and it depends on the experience of the user. I would prefer additional physical activity instead of a correction when ever possible.
Scenario 2: BG of 250 two hours after the meal. The IOB is now 2.5 because now only 70% of the original dosage is left. The proposed correction for 250 will be 3.5 and minus the IOB of 2.5 it will be 1. Here the proposed solution of the Glucosurfer has a good chance to lead to the target.
In general like you said it actually needs a simulation of insulin unfolding, digestion, food composition, insulin decomposition in the kidneys, receptor activity, muscle activity and so forth. This is why they all prefer the simple generalized IOB curve. On the other hand the simulation might suffer from the same generalizations. There are too many variables involved I am afraid. I also think that the IOB formula is a feasible approach - with a grain of salt.
I forgot to add that our IOB is calculated by taking the insulin both from carb and from correction boluses. This IOB is only accounted for in later corrections. This makes sense because for additional carbs you will eat later you will have to inject the normal dosage despite the IOB. I find it odd that some pumps only calculate the IOB from the corrections (how did they manage to get that approved?). After all physical activity can amplify the reaction of insulin and it does not matter if this insulin is for carbs or corrections.
I think, in theory, this is a great idea Holger.
However, What does this program, which will have to be personalized by a user anyway, get you that simply programming in an adjustable timer with an alarm won't?