The first data set that I decided to extract was related to a past post ( Predicting BG vs Predicting Sensitivity ) explaining how I decided to include sensitivity predicitons as part of the simpancreas algorithm.
The last simpancreas iteration which includes Endogenous Glucose Production calculations has been active on my loop since May, so I decided to wait for July to go by to at least show 3 months of new data.
Anyways, having much more sensitivity data now than when I first published that post has made me get a better view of what is going on with the data.
As always, when looking at this data, a couple of important questions popped up in my head:
1. How does sensitivity change as blood glucose changes with a positive / negative velocity?
2. How does sensitivity change as blood glucose changes with Carbohydrates on Board ( COB )?
3. How does sensitivity change as blood glucose changes with Carbohydrates on Board ( COB ) & positive / negative velocity?
I will try to go through a drill-down of the most interesting aspects I have observed with my insulin sensitivity and will try to answer these questions below using graphs.
PLEASE CONSIDER THE FOLLOWING:
All analysis, assumptions and conclusions are based on my own data and experimentation. They are not intended to show any professional statistical results whatsoever.
a) Graph showing average insulin sensitivities across a range of blood glucose levels ( 70mg/dl – 200mg/dl )
The interesting thing to notice here is insulin sensitivity seems to level-off on the extremes. It seems like the data is trying to tell me I have a max and min sensitivity.
Another interesting point for me is how sensitivity has a more pronounced s-shaped curve around the blood glucoses that make up my desired range ( 80mg/dl to 150mg/dl )
b) Graph showing average insulin sensitivities across a range of blood glucose levels ( 70mg/dl – 200mg/dl ) with and without Carbohydrates on Board
RED – No Carbohydrates on Board
BLUE – Carbohydrates on Board
In this graph notice how data representing “No Carbohydrates on Board” manifests as an s-shaped curve versus the “Carbohydrates on Board” which is a little bit more linear.
If you think about it, every time you consume carbs, you need extra insulin to reduce your blood glucose level vs the insulin you need when not consuming carbs.
This assumption is key to this post conclusion below…
c) Graph showing average insulin sensitivities across a range of blood glucose levels ( 70mg/dl – 200mg/dl ) WITH Carbohydrates on Board and variating velocity ( Positive vs Negative )
RED – Carbohydrates on Board
GREEN – Carbohydrates on Board and Negative Velocity
BLUE – Carbohydrates on Board and Positive Velocity
Yes, the ugliest of them all as I don’t have all the needed data for the selected blood glucose range.
If you think about it, its pretty hard and dangerous to get sensitivity data with Carbohydrates on Board with a NEGATIVE velocity if you are below 90 mg/dl.
A bit sad I can’t make that part of the analysis… but very happy that simpancreas has never put me on that situation.
d) Graph showing average insulin sensitivities across a range of blood glucose levels ( 70mg/dl – 200mg/dl ) WITHOUT Carbohydrates on Board and variating velocity ( Positive vs Negative )
RED – No Carbohydrates on Board
GREEN – No Carbohydrates on Board and Negative Velocity
BLUE – No Carbohydrates on Board and Positive Velocity
To me, this is the most interesting graph of them all. Why would insulin sensitivity variate this much ( about 10% in each direction ) when there is nothing to make it variate that much. This was something I thought about a lot before May 2016…
My hypothesis is that this variation can be attributed to Endogenous Glucose Production by the liver and my current iteration of code on simpancreas is using this data and assumptions.
At perceptus.org, we’re still working on the math behind EGP which will make more precise calculations… in the meantime…
What if the sensitivity differences when there are no carbs on board could be used to calculate how much glucose the liver is introducing or stopping to introduce to the bloodstream at any given time?
Comments are welcomed 🙂
UPDATE – July 11th, 2016
This is a preview of the upcoming data I will be publishing in August 2016 for blood glucose averages by month.
On the graph, my average blood glucose has been dropping throughout the year ( I have compared the data with last year and there are some similarities and some differences ) but it has done so more aggressively since I began using sensitivity analysis on the algorithm around March.
On May I did the latest updates to the calculations and I want to wait until July so I can have 3 months from March to May on first changes and then 3 months from May to July for latest changes and be able to compare them.