Dietary impact on blood glucose

After I moved to Paris I was getting a bit worried about the dietary changes that came with the move (eating all the baguettes with cheese that became my staple food, along with all the pastries). Triggered by this I got myself some freestyle libre glucose sensors, as those seemed to be an easy way to collect some data.

In total I’ve gone through 3 sensors so far (6 weeks of data). In addition to 4 weeks of my typical Parisian diet I also have 2 weeks of a changed diet (mainly consisting of ice cream, pastries overflowing with dulce de leche and small amounts of llama meat :llama::see_no_evil:).

While this makes an interesting natural experiment (for which I haven’t analyzed any data yet), I think I’d be interested in making more planned changes to my diet and see whether I can observe any effects. Michael said that he’d in principle would be interested in doing some self-experiments around blood glucose levels as well, so this could be a fun n=2 experiment opportunity.

Has anyone seen anyone else doing similar things in the past? Could be interesting to try similar approaches to have further references to compare to!

1 Like

We published a guide to self-research on blood sugar by Steven Jonas that may be useful:

I’ve tried some basic blood sugar testing, and presented it here:

In this project you’ll see that I had unexplained spikes at night. So I think if I participated in a group blood glucose testing project I might start with just trying to get a good baseline by tracking food and wearing a Freestyle Libre and see if there are still unexpected spikes.

Thanks for the links @Agaricus, that’s very valuable! For myself I did get some good baseline data during the last 6 weeks and I noticed some interesting spikes and also drops (especially during the night, probably when applying pressure to the sensor by sleeping on it).

But I agree, getting some baseline data in the group setting will be really valuable!

Measurement error was my first guess: was I sleeping on the sensor? But one reason I’m not sure is that the pattern is 4 hours long, and has a nice symmetrical shape across 5 hours. I’d think if I were sleeping on the sensor it would show the effects more suddenly and change when I moved. Will be interesting to try again!

Yeah, here’s an example of the nightly depressions in glucose that i attribute to me sleeping on the sensor, those come comparatively abrupt and then ‘normalize’ after a while.

Never was doing glucose measurements before, but I would be in checking the glucose level in the future, although trackers of glucose are not the easiest to have :grimacing:

1 Like

My colleagues and I recently published an n-of-1 trial we ran on ourselves examining sleep-deprivation effects on continuously monitored blood glucose (and 2 other outcomes). Our hope was that our study would help others design and analyze theirs in a similarly rigorous way. As such, I hope you may find it useful, @gedankenstuecke! Let me know if you’ve any questions.

“Effects of Sleep Deprivation on Blood Glucose, Food Cravings, and Affect in a Non-Diabetic: An N-of-1 Randomized Pilot Study”


Well, based on the fact that we’re all based in Paris we could do a group order of sensors for Michael, you & me as they are not that hard to get in Europe :smiley:

1 Like

Hey @ericjdaza, that’s super interesting and a great inspiration! I love the rigorous study design on the outset and hope we can do something similar for our self-research project!

1 Like

The Freestyle Libre sensors are notoriously variable – about 1 in 4 have failed for me, and the ranges tend to be different even on those that work. So you’ll need some sort of normalization before you can compare across people.

I put everything I know about CGM in a this doc.

Speaking of Paris and France, check out Jessie Inchauspie’s Instagram.


Thanks for the document! That’s awesome! One important thing one could add: Not only are the modalities for getting the sensors different in different countries, they are also geo-coded!

I tried to pair my freestyle libre sensor with the iPhone app to read out the data, and initially it claimed to not find any sensor! The problem was that my sensor was bought in France and my iPhone still ran on the US iTunes account, giving me the US version of their app!

The only way to get around this was to change my iTunes account to the German store (as that’s where I had a bank account available) and then download the German/EU version of the app. This led to the sensor being correctly identified and paired. So beware, sensors might not be compatible if you’re working around multiple countries!


I met up with @MichaelR this morning for breakfast and we had a chance to chat a bit about this self-research experiment. One simple but interesting question would be if/how the blood glucose response to the same kinds of food changes between and inside individuals.

For that we could prescribe the consumption of certain food items and monitor the blood glucose response that follows. By repeating the consumption over the monitoring period we could get replicates for the individual, allowing to see the variation for a single person. By having more than one person do the experiment, we could also see the variation between people.

@liubovv Does that sound like something you’d be interested in as well?

Next to-dos for doing this would be: Coming up with a protocol (which kinds of foods, how many repetitions, spacing of them, etc).

1 Like

Yes, this sounds good to me too.

But I would add here another question, why I am interested in this topic:
if I understand correctly there are different factors, which may influence glucose level decay (e.g. external factors not just food right?).

So I would be interested to study how one could dissect the influence from food and other factors on this decay.
Why this is interesting? It is known that people with type 1 or 2 diabetes under mental stress generally experience an increase in their blood glucose levels. When you’re under physical stress, your blood sugar can also increase. So question is what and when one could actually see?

Or this is a bit difficult? @gedankenstuecke
I see there could be some overlap with stress-level-app.
Which are the best apps for stress control? @MichaelR @gedankenstuecke

I think that could be interesting as well, same for the influence of physical activity etc. But it certainly would be adding complexity to the experiment. I wouldn’t know a good app for stress recording, but maybe @Agaricus has an idea!

There are an interesting set of questions. For a personal research project in which we hope to gain some immediate insight, I think food+glucose is most efficient approach. While there is doubtless some influence from stress, sleep, and other factors, the question is whether a clear pattern can be seen despite these influences. I’d be more likely to participate in a simple project like this where we focused on understanding the measurement protocol, building confidence through some calibration with finger-prick measurement, and running a few A-B-A-B trials with specific foods. While the questions we can answer with this approach are very simple, the benefit is high, because we develop confidence in our method. I think adding additional data types, especially ones with their own measurement challenges, such as stress, would be for a later stage.

Specifically on stress, in my stress tracking I’ve found it useful to track only “episodes of high stress.” By focusing on incidents of high stress I avoid the hard questions about calibration of my stress measures. (Is a “2” today the same as a “2” yesterday? Who knows?) Episodes of high stress are easily detected, and for me they are the phenomenon of key concern. I don’t really care about more subtle fluctuations in activation.

1 Like

While the questions we can answer with this approach are very simple, the benefit is high, because we develop confidence in our method. I think adding additional data types, especially ones with their own measurement challenges, such as stress, would be for a later stage.

Thanks, that was also my feeling. It’s probably good to start out simple for now and then complicate things at a later point once there’s some experience from the first monitoring!

If you live in Seattle, my friends and I have launched a new business to test this. We’ve partnered with a clinic (to give you the CGM prescription) and a food delivery service (to deliver precise, known quantities of fully-prepared food). “A 10 day clinical trial on yourself”. We’re not ready to publicize this widely, but I’d appreciate any feedback from OpenHumans friends.


That’s a really cool idea and I’d love to give that a try! Unfortunately all of us that are interested in giving it a try are based in Paris, so shipping ready-made food from Seattle isn’t an option. :joy:

Hi Richard, I’m with Bastian on this – the learning from blood glucose part is really interesting, both access to the device and, perhaps even more important, access to your help and knowledge. As feedback on the offering, I’d just say: that’s something I’d pay for! But the meal delivery and the associated cost puts it out of reach for me.

@liubovv @MichaelR I was pointed to this paper today by Guy:

Should definitely make an interesting read for us in the run-up to our own research!