This self-experiment is being done as part of the Keating Memorial Self-Research Project. A couple of other people from the Open Humans community are also running the same experiments. If you’re interested in joining in, let me know in the comments or send me a PM.
This post is an update on my experiments measuring the effect of food ingredients on blood sugar.
This is especially problematic for predicting the blood glucose impact of foods from their nutrition information as based on my data so far, even insoluble fibers can range in impact from 0.4 – 76% of glucose.
Next week I’ll finish out the major macronutrient groups with cornstarch. Still deciding where to go after that, but it will either be more ingredients used in low carb cooking (inulin, erythritol, soluble corn fiber, lupin flour) or mixtures of the major macronutrients (to measure combinations effects.
Details
Purpose
To quantify the effect of ingestion of food ingredients and ingredient combinations on my blood sugar.
Procedure. From 7 pm the day before through 4:30p the day of experiment, no food or calorie-containing drinks were consumed and no exercise was performed. Non-calorie-containing drinks were consumed as desired (water, caffeine-free tea, and decaffeinated coffee). At ~12 pm, the substance to be tested was dissolved or suspended in 475 mL of water and ingested as rapidly as comfortable. BGM measurements were then taken approximately every 15 min. for 2 h or until blood glucose had returned to baseline, whichever was longer. A final BGM measurement was taken 4.5 h after ingestion.
Measurements. Blood glucose was measured using a FreeStyle Libre flash glucose monitor and a FreeStyle Freedom Lite glucose meter with FreeStyle lancets & test strips. No special precautions were taken to clean the lancing site before measurement. To take a sample, the lancing devices was used to pierce the skin at an ~45 deg. angle from the finger. Blood was then squeezed out by running the thumb and pointer finger of the opposite hand from the first knuckle to the lancing site of the finger. Blood was then wicked into a test strip that had been inserted into the meter and the glucose reading was recorded.
Data Processing & Visualization. iAUC was calculated using the trapezoid method (see data spreadsheet for details). Data was visualized using Tableau.
Figure 1. Change in blood glucose vs. time for resistant starch and glucose tests.
Figure 2. Maximum blood glucose increase and iAUC vs. amount consumed. Red, orange, and blue indicate glucose, resistant wheat starch, and whey protein isolate, respectively. Lines are the best linear fit to the data.
Changes in blood glucose as a function of time for the resistant wheat starch and glucose tests are shown in Figure 1. For resistant wheat starch, I observe an increase in blood glucose starting at ~45 min. and reaching a peak between 75-120 min. While the timing is similar to that of whey protein, the magnitude of impact is much larger, with the peak change in blood glucose and iAUC increasing by 33% and 76% vs. glucose. It will be interesting to see next week how regular corn starch compares to resistant wheat starch and therefore if the chemical modifications to resist digestion are having any meaningful impact.
Put together, this indicates that resistant wheat starch is slower to digest than glucose, but contrary to the claims above, is clearly still metabolized to glucose. This is extremely disconcerting, as both oat fiber (iAUC 0.4% of glucose) and resistant wheat starch (iAUC 75% of glucose) are listed as insoluble fiber on nutrition labels, but have radically different impact on blood sugar. Given the lack of clarity and quantification of ingredient lists, this makes it nearly impossible to predict the blood glucose impact of a food without eating it and testing.
Next week I’ll finish out the major macronutrient groups with cornstarch. Still deciding where to go after that, but it will either be more ingredients used in low carb cooking (inulin, erythritol, soluble corn fiber, lupin flour) or mixtures of the major macronutrients (to measure combinations effects.
This self-experiment is being done as part of the Keating Memorial Self-Research Project. A couple of other people from the Open Humans community are also running the same experiments. If you’re interested in joining in, let me know in the comments or send me a PM.
This post is an update on my experiments measuring the effect of food ingredients on blood sugar.
This week, I have the results from whey protein and olive oil.
Summary
Olive oil had a negligible effect on my blood sugar, ~0.1 mg/dL/g(olive oil) for ~350 kcal of oil, or 1.5% that of glucose.
Whey protein isolate increases my blood sugar by ~20% that of glucose (by iAUC), but with a slower rise. This result sin a lower peak, 0.68 mg/dL/g(whey) or 10% that of glucose, but a long tail of increased blood sugar, ~0.4 mg/dL/g(whey) @ 4.5 h.
Still deciding what to try next, but it will either be corn starch (to have an example from each major macronutrient), resistant starch (fiber with disputed claims to non-digestibility), or combinations of protein, fat, or fiber with sugar.
Details
Purpose
To quantify the effect of ingestion of food ingredients and ingredient combinations on my blood sugar.
Ingredient Background
Whey protein isolate is a complete protein extracted from milk whey. It’s the most popular protein supplement due to its ease of digestion, rapid absorption, and appreciable content of all 9 essential amino acids.
Olive oil is a cooking oil that’s high in unsaturated fats, primarily oleic, linoleic, and palmitic acid. It’s used extensively in cooking.
Design/Methods
Procedure. From 7 pm the day before through 4:30p the day of experiment, no food or calorie-containing drinks were consumed and no exercise was performed. Non-calorie-containing drinks were consumed as desired (water, caffeine-free tea, and decaffeinated coffee). At ~12 pm, the substance to be tested was dissolved or suspended in 475 mL of water and ingested as rapidly as comfortable. BGM measurements were then taken approximately every 15 min. for 2 h or until blood glucose had returned to baseline, whichever was longer. A final BGM measurement was taken 4.5 h after ingestion.
Measurements. Blood glucose was measured using a FreeStyle Libre flash glucose monitor and a FreeStyle Freedom Lite glucose meter with FreeStyle lancets & test strips. No special precautions were taken to clean the lancing site before measurement. To take a sample, the lancing devices was used to pierce the skin at an ~45 deg. angle from the finger. Blood was then squeezed out by running the thumb and pointer finger of the opposite hand from the first knuckle to the lancing site of the finger. Blood was then wicked into a test strip that had been inserted into the meter and the glucose reading was recorded.
Data Processing & Visualization. iAUC was calculated using the trapezoid method (see data spreadsheet for details). Data was visualized using Tableau.
Figure 1. Change in blood glucose vs. time for whey protein and olive oil tests.
Figure 2. Change in blood glucose vs. time for whey protein and glucose for conditions with a peak of 20-25 Δmg/dL.
Figure 3. Maximum blood glucose increase and iAUC vs. amount consumed. Red, blue, and green indicate glucose, whey protein isolate, and olive oil, respectively. Lines are the best linear fit to the data.
Changes in blood glucose as a function of time for the whey protein isolate and olive oil tests are shown in Figure 1. As expected, olive oil showed no measurable impact on blood glucose at 40 g, or 350 kCal, consumed. Future experiments will look at whether it or similar oils can modulate the blood sugar response to ingredients that do impact blood sugar.
For whey protein, I observe an increase in blood glucose starting at ~45 min. and reaching a peak between 75-105 min. The magnitude increases with increasing amount consumed, but non-linearly; the difference between the 15 and 30 g consumed conditions is quite small. More data is needed at lower amounts consumed to see if this is a real effect or just noise in the data.
Comparing directly to glucose, for the same peak change in blood glucose, whey protein is much slower to impact my blood glucose and is metabolized over a much longer period of time. For example, looking at the conditions where peak Δmg/dL = 20-25 (see Figure 2):
Time to >5 mg/dL rise is 60 vs. 15 min. for whey vs. glucose
Time to return to <5 Δmg/dL is 255 vs. 120 min. for whey vs. glucose
Results are similar for all other amounts consumed. As show in Figure 3 and the summary table, this slower metabolism results in whey protein having a larger relative impact on iAUC than peak change in blood glucose (20 vs. 10% of glucose per gram). This may be do to giving my body more time to produce endogenous insulin, or even directly stimulating its production, reducing the peak blood glucose. Both of these effects have been reported. Given that, it would be useful to see the same measurements in someone with Type 1 diabetes, who does not produce endogenous insulin.
Conclusion & Next Experiments
Olive oil had a negligible effect on my blood sugar, ~0.1 mg/dL/g(olive oil) for ~350 kcal of oil, or 1.5% that of glucose.
Whey protein isolate increases my blood sugar by ~20% that of glucose (by iAUC), but with a slower rise. This result sin a lower peak, 0.68 mg/dL/g(whey) or 10% that of glucose, but a long tail of increased blood sugar, ~0.4 mg/dL/g(whey) @ 4.5 h.
Still deciding what to try next week, but it will either be corn starch (to have an example from each major macronutrient), resistant starch (fiber with disputed claims to non-digestibility), or combinations of protein, fat, or fiber with sugar.
This self-experiment is being done as part of the Keating Memorial Self-Research Project. A couple of other people from the Open Humans community are also running the same experiments. If you’re interested in joining in, let me know in the comments or send me a PM.
This post is an update on my experiments measuring the effect of food ingredients on blood sugar.
This week, I have the results from oat fiber and got started on whey protein.
Summary
Oat fiber has a negligible effect on my blood sugar, <0.05 mg/dL/g(oat fiber), or <0.5% that of glucose. Cooking the oat fiber had no significant effect. So, I’m safe to keep baking with it ☺.
Details
Purpose
To quantify the effect of ingestion of food ingredients and ingredient combinations on my blood sugar.
Procedure. From 7 pm the day before through 4:30p the day of experiment, no food or calorie-containing drinks were consumed and no exercise was performed. Non-calorie-containing drinks were consumed as desired (water, caffeine-free tea, and decaffeinated coffee). At ~12 pm, the substance to be tested was dissolved or suspended in 475 mL of water and ingested as rapidly as comfortable. For the cooked oat fiber, 100 g oat fiber was mixed with 200 g water and allowed to hydrate for 30 min. The mixture was then baked in a parchment lined muffin tin for 18 min. at 350 °F. The oat fiber was then suspended as described above. BGM measurements were then taken approximately every 15 min. for 2 h or until blood glucose had returned to baseline, whichever was longer. A final BGM measurement was taken 4.5 h after ingestion.
Measurements. Blood glucose was measured using a FreeStyle Libre flash glucose monitor and a FreeStyle Freedom Lite glucose meter with FreeStyle lancets & test strips. No special precautions were taken to clean the lancing site before measurement. To take a sample, the lancing devices was used to pierce the skin at an ~45 deg. angle from the finger. Blood was then squeezed out by running the thumb and pointer finger of the opposite hand from the first knuckle to the lancing site of the finger. Blood was then wicked into a test strip that had been inserted into the meter and the glucose reading was recorded.
Data Processing & Visualization. iAUC was calculated using the trapezoid method (see data spreadsheet for details). Data was visualized using Tableau.
Figure 1. Change in blood glucose vs. time for oat fiber tests.
Change in blood glucose as a function of time for the oat fiber tests is shown in Figure 1. Qualitatively, there appears to be no impact of oat fiber up to 100 g consumed. It was extremely uncomfortable to drink that much oat fiber in one sitting, so it’s unlikely I will ever eat more than that.
Figure 2. Maximum blood glucose increase and iAUC vs. amount consumed. Red, blue, orange, and green indicate glucose, allulose, oat fiber, and cooked oat fiber, respectively. The line is the best linear fit to the data.
To better quantify the impact of glucose on my blood glucose, I plotted the maximum increase in blood glucose and the iAUC of blood glucose (incremental area under the curve) vs. amount consumed for all ingredients tested so far (see Figure 2). While the oat fiber data shows an increase in both blood glucose and iAUC as a function of amount consumed, there’s only two data points and the magnitude is extremely small and could easily be due to experimental error. Confirming this effect would require running more measurements. I may go back and do this later, but for the moment, I would prefer to focus my time on ingredients with unknown or more substantial effects.
Since oat fiber is used in baking, I also wanted to check if heating it would break down the fibers and increase digestibility. Towards that end, I mixed 100 g of oat fiber with 200 g of water and baked at 350 °F for 18 min. (time & temperature for my muffin recipe and longer than my cookie recipe), then suspended it in water using the same procedure as with the uncooked fiber. There was no observable change in appearance, taste, or texture from the cooking process. There was a very minor increase in blood glucose response, but well within the measurement error of the meter.
Conclusion & Next Experiments
Oat fiber has a negligible effect on my blood sugar, <0.05 mg/dL/g(oat fiber), or <0.5% that of glucose. Cooking the oat fiber had no significant effect. This week, I will measure the effect of whey protein, a common protein supplement.
This post is the final report on our Community Self-Experiment studying the effect of hot showers on blood glucose. If you don’t want to read all the details, the highlights are in the Background & Summary section immediately below.
To figure out what’s really going on, we decided to do a communal self-experiment. 8 Redditors with diabetes developed an experimental protocol, measured their blood glucose before and after 41 showers using a combination of CGMs and BGMs, and analyzed the results.
Summary of Results
By working together, the team of experimenters was able to learn more and learn faster than any one of us would have been able to on our own. From the data, we were able to answer several of our initial questions:
What is the change in blood glucose after a hot shower under controlled conditions?
From BGM: 12 ± 17 mg/dL
From CGM: 21 ± 15 mg/dL
Is the observed change in blood glucose real or a CGM sensor artifact?
The change is real, not a sensor artifact (change is observed with BGM; CGM measurements are consistent with typical variation between CGM and BGM)
We cannot rule out the difference in effect size between BGM and CGM being due to a sensor artifact, but the data does not provide support for this hypothesis.
Is there significant person-to-person variation in the magnitude or direction of the effect?
The data is consistent with person-to-person variation, but significantly more data and/or more controlled conditions are required to determine if the variation exists and it’s effect size.
Is the change in blood glucose cause by the hot shower?
We can not identify any factors that account for the blood glucose rise other than the shower, but would need significantly more data and/or more controlled conditions to be certain that the shower is the cause.
While we were not able to get firm answers to all of our questions, we did get a measure of the effect size and rule out it being a CGM sensor artifact (the leading hypothesis in the original post). We also learned a lot that will help guide future Community Self-Experiments.
Data Processing. All data was converted into consistent units and put into an excel spreadsheet. From the raw data, I calculated change in BG from start of shower, as well as the largest relative change, and the time until largest relative change (see spreadsheet for calculation details). Visualization was done using Tableau.
What is the change in blood glucose after a hot shower under controlled conditions?
To answer this question, I plotted largest observed change over the 1 hour monitoring period for each shower as measured by both BGM and CGM (see Figure 1).
Figure 1. Max ΔBGM & ΔCGM for each shower, colored by experimenter. Reference band shows average +/- 1 standard deviation.
Looking at the data in Figure 1:
There is a large rise in blood glucose following a hot shower, though with significant variance in the size of the effect.
The rise is observed for both BGM (12 ± 17 mg/dL) and CGM (21 ± 15 mg/dL) measurements.
By count, we see (1 measurement excluded due to recording error):
>5 mg/dL increase: 34/40 (85%)
>5 mg/dL decrease: 3/40 (7.5%)
<5 mg/dL change: 3/40 (7.5%)
Conclusion: Blood glucose showed a consistent, measurable increase within 1h of taking a hot shower.
Is the observed change in blood glucose real or a CGM sensor artifact?
Looking again at Figure 1, the increase in blood glucose is seen for both BGM and CGM measurements, indicating that it can’t be just a CGM artifact.
To further confirm this conclusion, we looked at the data from person H comparing BGM vs. CGM measurements during the normal course of the day vs. after a shower. As shown in Figure 2, for a single Libre sensor, there is a linear relationship between measured blood glucose by BGM vs. CGM and the data collected immediately and 15 minutes after a shower mostly lies within the normal variance in the data, with all exceptions showing a lower blood glucose measured by CGM. This indicates that any variation in CGM data due to a sensor artifact is smaller than the observed increase in blood glucose. Note that while this confirms that the measured effect is not exclusively due to a sensor artifact, it is still possible that a sensor artifact accounts for the difference in effect size as measured by BGM vs. CGM (12 vs. 21 mg/dL).
Figure 2. Blood glucose measured by FreeStyle Libre and FreeStyle Freedom Lite for person H over the course of 10 days. Grey line is a linear fit to the data and data collected immediately and 15 min. after a hot shower is shown in red.
Conclusion: The observed increase in blood glucose is not a CGM sensor artifact (though a partial effect from the CGM sensor is not ruled out).
Is there significant person-to-person variation in the magnitude or direction of the effect?
Looking again at the data in Figure 1:
A >5 mg/dL increase in blood sugar is observed for 6/8 (75%) of participants, with 2/8 (25%) showing a >5 mg/dL decrease in blood sugar.
Only 2 participants provided multiple measurements, A and H. For those we observe:
A: 12 ± 16 mg/dL
H: 26 ± 14 mg/dL
The difference is statistically significant (Welch’s t-test, p=0.016), but since the measurements were made using different methods (CGM for A, BGM for H), times (10 min. for A, 20 min. for H), and temperatures, this is only weak evidence for person-to-person variation.
Conclusion: The data is consistent with person-to-person variation, but significantly more data and/or more controlled conditions are required to determine if the variation exists and it’s effect size.
Is the change in blood glucose cause by the hot shower?
This is the most difficult question to answer. In hindsight, we should have done some randomized experiments where the experimenters held conditions as constant as possible, randomly decided whether or not to shower, and measured blood glucose either way. In the absence of that data, we analyzed the data we had for any correlation between the blood glucose rise and non-shower factors. It should be noted that the protocol did not control for any of these factors, so no causation or lack thereof should be inferred from the analysis.
Max ΔBGM or Max ΔCGM vs. hour of the day – no trend across the whole data set, nor within experimenters
Max ΔBGM vs. starting BGM – no trend across the whole data set, but within Experimenter H’s data, there’s an indication of a negative correlation (R2 = 0.32, p = 0.045).
Max ΔCGM vs. starting CGM – no clear trend across the whole data set, nor within experimenters.
Max ΔBGM vs. Temperature – no clear trend across the whole data set, nor within experimenters. Note: most experimenters did not record the shower temperature and the one who did (Person H) kept the temperature within ±3 °C.
Max ΔBGM or Max ΔCGM vs. Time since last meal or medication – There’s a positive correlation over the whole data set, but it doesn’t hold up within the two experimenters with repeat measurements, suggesting that it’s an effect person-to-person variation, possibly caused by systematic variation in conditions.
Conclusion: We can not identify any factors that account for the blood glucose rise other than the shower, but would need significantly more data and/or more controlled conditions to be certain that the shower is the cause.
Conclusions & Lessons Learned
By working together, the team of experimenters was able to learn more and learn faster than any one of us would have been able to on our own. From the data, we were able to answer several of our initial questions.
Conclusions:
What is the change in blood glucose after a hot shower under controlled conditions?
From BGM: 12 ± 17 mg/dL
From CGM: 21 ± 15 mg/dL
Is the observed change in blood glucose real or a CGM sensor artifact?
The change is real, not a sensor artifact (change is observed with BGM; CGM measurements are consistent with typical variation between CGM and BGM)
We cannot rule out the difference in effect size between BGM and CGM being due to a sensor artifact, but the data does not provide support for this hypothesis.
Is there significant person-to-person variation in the magnitude or direction of the effect?
The data is consistent with person-to-person variation, but significantly more data and/or more controlled conditions are required to determine if the variation exists and it’s effect size.
Is the change in blood glucose cause by the hot shower?
We can not identify any factors that account for the blood glucose rise other than the shower, but would need significantly more data and/or more controlled conditions to be certain that the shower is the cause.
While we were not able to get firm answers to all of our questions, we did get a measure of the effect size and rule out it being a CGM sensor artifact (the leading hypothesis in the original post). We also learned a lot that will help guide future Community Self-Experiments.
Key Lessons Learned:
Community Self-Experiments enable collection of data much faster than single-person experiments, both because more people are collecting data and because the group activity motivates participants.
Take more care with the experimental design, especially the implementation of control experiments to help rule out alternate hypotheses.
Implement better data sharing/management. In this experiment, data was posted, then manually entered into an excel sheet, which was very time consuming.
This self-experiment is being done as part of the Keating Memorial Self-Research Project. A couple of other people from the Open Humans community are also running the same experiments. If you’re interested in joining in, let me know in the comments or send me a PM.
This post is an update on my experiments measuring the effect of food ingredients on blood sugar.
Procedure. From 7 pm the day before through 4:30p the day of experiment, no food or calorie-containing drinks were consumed and no exercise was performed. Non-calorie-containing drinks were consumed as desired (water, caffeine-free tea, and decaffeinated coffee). At ~12 pm, the substance to be tested was dissolved or suspended in 475 mL of water and ingested as rapidly as comfortable. BGM measurements were then taken approximately every 15 min. for 2 h or until blood glucose had returned to baseline, whichever was longer. A final BGM measurement was taken 4.5 h after ingestion.
Measurements. Blood glucose was measured using a FreeStyle Libre flash glucose monitor and a FreeStyle Freedom Lite glucose meter with FreeStyle lancets & test strips. No special precautions were taken to clean the lancing site before measurement. To take a sample, the lancing devices was used to pierce the skin at an ~45 deg. angle from the finger. Blood was then squeezed out by running the thumb and pointer finger of the opposite hand from the first knuckle to the lancing site of the finger. Blood was then wicked into a test strip that had been inserted into the meter and the glucose reading was recorded.
Data Processing & Visualization. iAUC was calculated using the trapezoid method (see data spreadsheet for details). Data was visualized using Tableau.
Figure 1. Change in blood glucose vs. time for alluose tests.
Change in blood glucose as a function of time for the allulose tests is shown in Figure 1. Qualitatively, there appears to be no impact of allulose up to 60 g consumed, with the possible exception of a small around 75 min.
Figure 2. Maximum blood glucose increase and iAUC vs. amount consumed. Red and blue indicate glucose and allulose, respectively. The line is the best linear fit to the data.
To better quantify the impact of glucose on my blood glucose, I plotted the maximum increase in blood glucose and the iAUC of blood glucose (incremental area under the curve) vs. amount consumed for both glucose and allulose (see Figure 2). While the allulose data shows an increase in both blood glucose and iAUC as a function of amount consumed, there’s only two data points and the magnitude is extremely small and could easily be due to experimental error. Confirming this effect would require running more allulose measurements. I may go back and do this later, but for the moment, I would prefer to focus my time on ingredients with unknown or more substantial effects.
Lastly, I continue to observe a large negative intercept, suggesting a background drop in blood sugar during the experimental window.
Conclusion & Next Experiments
Allulose has a negligible effect on my blood sugar, <0.1 mg/dL/g(allulose), or <2% that of glucose. This week, I will measure the effect of oat fiber, a zero calorie, zero digestible carbohydrate flour replacement.
