Insulin dosage calculation

I respectfully suggest that part of your problem is that you are not injecting enough. This disease isn’t only about calculations, if it was there would be no need for closed loop pumps. Part of the reason so many people struggle, especially those on MDI, is that they are taught they just need to take 1 dose of insulin according to a carb ratio prior to any meal and they should be in range 2 hours later. You’ve already proven through your experience that this is impossible. It’s not a matter of calculating better, it’s because food doesn’t match the curve of the insulin perfectly along with a million other variables affecting absorption, digestion, destruction of injected insulin by the immune system, etc. Plus when I said multiple injections during meals I was also talking about when I am eating something and then decide to eat dessert or something else and have to inject again. This is what people on pumps do and I’m just as happy to pull out my pen every time I want to eat more, but my point was I don’t want to have to pull out my phone and log/review data every time I do this.

2 Likes

you hit the nail on the head with that answer.

1 Like

Eric

I do inject enough as I continuously inject with a pump. I agree with what you are saying especially the pre-set ratio’s used in my pump(s) as they are never right based a body constantly changing. Ratio’s don’t work with a constantly changing metabolism and a Vargas nerve that makes up its own mind with my gastro paresis. I can do exactly the same food (I’m not talking carbs I’m talking food) at the same times two or three days in a row with the same exact liquids intake and all 3 days my BG’s are completely different.

My body doesn’t and never has reacted to insulin per a curve as indicated by the manufacturer. I currently use U500 in my pump and the 2-3 hour curve is more likely a 5-6 hr curve in regards to bolus(injection) reaction. I had no difference with other fast acting insulin (such as Apidra) or for that matter with long acting ones for basal in the pump and shot with short acting for bolus.

Yes I gave up logging and keeping track a long time ago as it didn’t make a difference and after 50+ years with T1, I’m relatively healthy especially compared to my peers.

I probably (or definitely) won’t live far into my 90’s but I have lived good and T1 hasn’t ruled it.

I’m happy that pens work for you, they never worked for me although I thought they were a 100x better than the syringes & vials.

Stay healthy

2 Likes

I think at the end of the day I’m with TomWu on this. The engineer in me really wants to precisely measure the carb/protein/fat I ingest and precisely time the insulin boluses (normally a lot of them every day) I deliver to compensate, but the T1D knows it does not work.

It isn’t a weather prediction problem; insulin bolusing is inherently soluble insofar as it is possible to determine the amount of insulin required in our veins (not our skin) based on what we ate, what the ambient temperature is, etc. But forget that; we do not and cannot know what we eat, particularly in the US where food labels are a sad joke if they are present at all.

So, yes, I very much favor an approach that does what I termed “reverse engineering”; I think I didn’t mean what TomWu understood but that. I meant that given a sequence of CGM readings it is possible to determine an appropriate insulin bolus without any other information. The bolus will be late, but that is actually what we all do; the immediately preceding comments in this thread demonstrate that.

John Bowler

2 Likes

I don’t buy any pre-prepared food and measure everything I eat to the gram, so that I consume exactly the same meal every day at exactly the same time. I have done this now every single day for many, many years. There cannot really be much variation in the demand for insulin there. My activity levels, largely because I don’t exercise and live by a pretty fixed routine, is as near to identical every day as possible. Despite this, blood sugar levels and insulin requirements are all over the place. Just in the last few days my total daily insulin requirements have risen by more than 30%, but from where or what? I know from experience that whenever there is a sudden cold snap, my blood sugar spikes, and since there was a sudden cold snap earlier this week, that was it.

My job involves me lecturing to students, and my blood sugar, depending on the difficulty, controversy, and opposition that the lecture arouses in the students, can rise by as much as 80 points, or not at all, if everything is just routine. But I can’t plug ‘routine lecture’ or ‘stressful lecture’ into a blood sugar regulating machine and base the dose on that, because the stress effect of anything on blood sugar is utterly unpredictable.

Even if I have stressful dreams I will wake up with higher than normal blood sugar, but how on earth do I anticipate what I am going to dream?

With every single measurable blood sugar effect kept constant, as it has been for years, here are the pre-dinner blood sugar levels for two weeks ago (expressed on the scale of 5 equals the physiological level):

Day One: 3.9
Day Two: 4.6
Day Three: 10.6
Day Four: 9.7
Day Five: 4.0
Day Six: 4.0
Day Seven: 3.5

The only possible response to this is that based on the variables I can and do control there is no way to anticipate what is going to happen; there are more hormones operative in the human body than have even been identified yet, and many of them affect blood sugar, from adrenalin to cortisol to thyroid, so it’s just going to be unpredictable without functioning beta cells.

2 Likes

That’s an interesting response, because you are refuting the base engineering approach of measuring everything and yet not suggesting that it is a weather prediction problem; although that is a possible, reasonable, conclusion given your statement I have yet to be unable to explain what went wrong in my own case. I do completely accept the stress issue; I’ve spent my life quite consciously avoiding it (stress me when I’m working and I won’t work for, or with, you any more).

So are you part of the diafyt MedTech operation?

Such an approach is clearly the potential solution to the very real problems you identify.

John Bowler

I personally would be fine with trying out this app or something like it, BUT I am a pretty skeptical diabetic and wouldn’t want to purchase it until I could thoroughly try it out. I wouldn’t have any problem inputting data since I already do that using the OneDrop and Spike apps and I don’t find it to be tedious. However if all this is is a insulin calculator I’m not sure how much I’d actually use it since I rarely count carbs since I’m on a low carb diet and don’t usually need to.

I think they are great. I use a Google Sheet that I created to calculate for need but of course I have to enter all the data. I have looked at other Android apps but they are too complicated. If it works, use it.

I’m a T1D myself as probably as same as skeptical as you are … so yes, it took a while to get used to it. The diafyt app gives you there calculated recommendation that you can always overwrite (what I did mostly in the beginning). However I realized that in many cases the calculated dosage amount would have been the better choice so I followed the advise more often. Now, after about 8 weeks I follow the given advise to about 90%.
The app is super simplified a picture is here: https://twitter.com/diafyt/status/1049403169401425920

may I ask … are you type one or two?

1 Like

I’m part of the diafyt team. I’m a type one and doing all testing. I’m also a researcher doing algorithm design and tuning.

I can give you an idea of what we are doing all day: Machine learning is quite tricky especially in such a complicated case like diabetes. For example we added lots of complications to our calculation e.g. injection location, stress, weather, and more. Machine learning looks for relevance and kills everything what is not relevant after a while and surprisingly in most cases all these complications went away. … So for my part I keep busy researching what is relevant.

… great answer.

Yes, the insulin calculation we are performing is more than just multiplication and data entry what most people think we do. There are 2 tools we are using

  1. algorithm: does simulate the diabetic metabolic system
  2. machine learning for parameter tuning

for example machine learning will search for relevance in your data and personal habits and tune the algorithm parameterization.

This takes some computing power so we placed it on Amazon Cloud to use there machine learning tools, reliability and performance for calculation. On your smartphone sits a simplified tool that keeps operating if you do not have internet access.

in cases like your cold strikes in, the algorithm will take some time to adapt but will regulate accordingly after a couple of hours.

for myself glucose levels have been largely unpredictable, also because of my lifestyle. I never eat the same, sleep the same, do sports the same … so my glucose levels have been between as low as 2.5 and as high as 25. It became better after using a CGM but still sometimes unpredictable. So with diafyt most of these described high and lows went away with my BGL in range for most of the time. Another benefit is that I do not need to do extra boluses.

ha … that sounds like my story. My high A1C caused my doctor to say that I will not reach my 65th birthday if I continue like that. So I guess I tried to become a pro in diabetes management however with limited success. About 4 years ago we started with diafyt. We spend 1 year just analyzing existing algorithms and loop systems to find out what makes them not working. After 2 more years of research we received EU funding and started building diafyt. Since about 3 months I have it on my Samsung and the results are surprisingly good. these frequent high and lows are all gone. And I did not change anything of my lifestyle. So for me it does work. My A1C is about 6.8 down from above 8. So Im happy and so is my doctor :slight_smile:

Lets give this discussion one more thought.

We at diafyt believe that optimal insulin calculation has the most relevance for diabetes management. Meaning the correct dosage would be all you need to manage your diabetes.

ok , that’s easier sad than done but it is where we at diafyt started. So what do YOU think. I appreciate your responses.

IMHO a statement like this makes you come across as somebody on the fringe and not to be taken seriously.

How long have you been a diabetic Tom?

1 Like

Type 1 for 26 of my 28 years. I think whether or not this type of app would work for me is how accurate I would need to be in things like carb counting, and will this app have calculations for Afrezza and older insulins like R?

I don’t think I understand this, unless you are using Regular/Actrapid insulin in addition to rapid insulin and dosing it perfectly, never eating anything that digests after the peak of rapid insulin, or (and not trying to judge in any way but just saying this because you put it in issue and are promoting this software) your A1C may be 6.8 and not better because you do need additional boluses to account for delayed digestion/protein spikes. An example of a meal I ate on the weekend is a steak and ice cream cake for dessert at a birthday. Even with using Regular insulin it still took several shots hours later to deal with the delayed spikes. This was not miscalculation, I was in range 3-4 hours after eating. It was after that when I started to rise and there is no perfect dose of insulin that could be taken before because the insulins we have don’t last long enough and peak too early causing lows.

3 Likes

lol - I suspect you do understand it. Perfectly.
:slight_smile:

One can only educate another person who is willing to learn.

1 Like

Big problem for me. Spiking late then crashing from 3 extra units of humalog now taking up to 2 or 3 bouls at 1 unit and leveling at 150
How do you figure that into an algarythin

This is what the AP/loop systems excel at, so I think it is possible to design a system to better deal with these issues (though not perfectly) as they happen by delivering more or less insulin on a minute-by-minute basis based on CGM readings. The snake oil here is the promise of a system that calculates everything so perfectly you can minimize the amount of injections needed!

1 Like