Tuesday, July 28, 2015

Freestyle Libre - non type 1 diabetic run - "normal" profile.

I've been sleeping on this blog for a while and this post is probably something like 7 months late. I couldn't resist running a Libre sensor on myself in December 2014 and had promised myself that I would release the data in full.

Personal Data

Adult Male, 51 years old. My height is 186 cm. My weight for the last two years has been oscillating between 90 and 97 kg, depending mostly on my level of physical activity which itself mostly depends on the quality of sleep I get. My conventional BMI varies from 26 to 28. Overweight, but not dramatically so (as I keep telling myself to avoid suicidal thoughts). My blood pressure is constantly 110/70 mm Hg (now, that is a good news). I have no family history of type 2 diabetes. I've recently had an abdominal CT scan with contrast for what turned out to be a sudden extremely severe lactose intolerance and don't seem to have significant visceral fat deposits (which would be a risk factor for type 2 diabetes). That CT scan showed a congenital Bochdalek hernia which I was unaware of (I had to look up that one after talking with the radiologist). My recent blood tests have been relatively normal, triglycerides close to the upper limit of normal range. I do, on a regular basis, suffer from bouts of relatively severe joint pain and, on those occasions, have tested with very high level of non specific auto-immune complexes. That summarizes the "patient" data.

Libre data

The data set of my Libre run can be downloaded here if you are interested. Please do note that my sleep and meal schedule is somewhat irregular as I don't typically have timing constraints. And here is the graphical overview of my 14 days run.

My average BG for the period was a clean 100 mg/dL. My standard deviation was 22 mg/dL. (AVG: 100.230 SD: 22.041). Excellent weeks with my T1D son can match the 100 mg/dL average, but there is still a 'cost' in standard deviation which we haven't been able to get below 33 mg/dL on a normal high carb diet.

One interesting point from the Libre perspective is that, as many T1D users have noticed, the Libre underestimated my ISIG during the 10 hours following insertion. That seems to be the standard behavior of the Abbott sensor, so typical that many users now pre-insert to avoid false low readings during the start-up period. There is no way a "normal" patient who is not being treated with insulin could spend hours in the 40 mg/dL range.

Of course, I immediately decided to stress test the system by eating a huge carb-only breakfast 12 hours after insertion and the Libre picked up the sudden BG increase quite nicely. The way I climbed after that meal had me worried: I would describe some of those spikes as possibly indicative of a slow phase 1 insulin response. This being said, for the sake of (amateur) science, I did push the envelope with carbs on some occasions. In all cases, my ISIG corrected itself quickly enough and I never woke up above 100 mg/dL. Strictly speaking, I don't fall either in the somewhat vague pre-Type 2 criteria or the more well defined Type 2 criteria. Still, as a somewhat health conscious individual, I feel I should monitor my Insulin Phase 1 response again when the sensor availability issue improves. 

Libre accuracy and behavior

I did run a lot of tests and attempted a few CGM like sessions. That means that in some cases I did up to 30 BG meter tests in a couple of hours. Including all the BG/Libre value pairs taken during test sessions, fast increases after test meals, fast drops after intense exercise or long 'raw data' flat sessions at my desk... would, without any doubt, bias the accuracy results. 

Here is the accuracy data for standard (outside any experiment or intensive testing session) BG meter tests.

and for our mmol friends

As you can see, the Libre seemed to under estimate values a bit on average. But, obviously, we should not derive any general rule from a single sensor, on a single person, with a low number of individual values. (yes, I am aware that some highly publicized medical studies have been run on 6 patients split in two groups of 3...)

Here's another view of the data.
Looking at the whole set of data (more than 200 BG comparison tests) I was under the impression that the Libre tracked lower and lower over time, particularly during day 12, 13 and 14. 

Other findings

There's an additional finding I would like to mention. As far as I am concerned, this is the most important lesson I learned from that Libre run. As I said above, I experimented quite a bit with my meals. In addition to a few carb orgies, I tested rice extensively. Here's the theory in terms of glycemic Index and glycemic load for a 150 grams serving

White rice, average8915043
Quick cooking white basmati6715028
Brown rice, average5015016
Converted, white rice (Uncle Ben’s®)3815014
Well, as far as my testing was concerned, I could not tell the difference between white, brown, converted white or converted brown rice. All types of commonly available rices behaved identically in terms of glycemic response. 

I already had the impression that the type of rice included in my son's meals did not matter and that they all triggered a response that was much closer to "fast carbs" than to "slow carbs" even when, in theory, they should have behave differently. Of course, with my son, there was always the uncertainty surrounding the IOB and the timing of his injections. In my case, this was not a factor...

Towards the end of the testing period I managed to find a brown rice - the type of extremely raw brown rice you purchase in bags and you slow cook for 30 minutes or more - that behaved as expected. Again, an issue that certainly deserves further investigations. Again an issue where I will not claim statistical significance... but my impression was that all pre-processed rice behaved as fast carb.

This is certainly something I'd like to investigate more in depth when the opportunity arises. Until then, enjoy your food!