Wednesday, March 30, 2011

Total Daily Insulin Versus Cycling Time

Thanks to my super-fantastic coach and team director Kori, I have been logging my workouts since she started working with me late last year.  From my insulin pump, I can also pull off my total daily insulin amounts (or "TDD" for total daily dose) since the beginning of time, give or take.  I thought it might be fun to just plot out my TDD as a function of my cycling time, even though the results are probably as shocking as showing that washing hands reduces the spread of colds.  Anyway, I decided to just do a very simple linear regression of the data between January and March, 2011.  This model ignores variables such as what other exercise I did that day (I often walk about 40 minutes per day), whether I ate more or less than normal, my weight, how old my infusion set was, etc.  In short, the only variable I am considering is time on the bike.  But there still is a clear trend that shows with increasing time on the bike, I took less insulin.

An Overly Simplistic Model But Still Sort of Interesting

For those who like details, the slope was -0.046 U/min, the y-intercept was 35.3 U and correlation coefficient was -0.65.  If I included only March, rather than January through March, the correlation coefficient was -0.68.  The mean TDD for January was 32 U and was 30 U for March.  Total cycling time was only 90 minutes more in March although the intensity was higher.  My weight was about 5 pounds less in March.  Again, one of the most important and obvious variables affecting TDD is calorie intake, which is not included here.  But still it was fun to plot this out since the data was readily available.   Maybe I will be better about logging total activity time, including other forms of exercise, for a period of time.  If I am super motivated, I might even track calories although I can only stand doing that once in a while.


Kevin said...

Fascinating stuff!

I don't keep track of my minutes of biking (mostly because there's not much variation - either I'm commuting or not), but I totally notice a drop in my insulin requirements on the day's I'm riding.

So, what strikes me about your data is the amount of variation you have when you ride 0 hours and 50 minutes. I'm guessing those 50 minute ride days are commutes? They don't seem to have any effect on your TDD.

I'm guessing that if you dropped those data from your analysis, you'd see a steeper slope. But as is, every 20 min of exercise appears to lower you TDD by 1 unit. Cool stuff to know!

Robert said...

What if you plot against TSS (or some other intensity-weighted summary, like BikeScore) rather than time?