Monday, January 30, 2017

Blood glucose variations in normal individuals: A chaotic mess

I love statistics. But statistics is the science that will tell you that each person in a group of 20 people ate half a chicken per week over six months, until you realize that 10 died because they ate nothing while the other 10 ate a full chicken every week.

Statistics is the science that will tell you that there is an “association” between these two variables: my weight from 1 to 20 years of age, and the price of gasoline during that period. These two variables are indeed highly correlated, by neither has influenced the other in any way.

This is why I often like to see the underlying numbers when I am told that such and such health measure on average is this or that, or that this or that disease is associated with elevated consumption of whatever. Statistical results must be interpreted carefully. Lying with statistics is very easy.

A case in point is that of blood glucose variations among normal individuals. Try plotting them on graphs. What do you see? A chaotic mess, even when the individuals are pre-screened to exclude anybody with blood glucose abnormalities that would even hint at pre-diabetes. You see wild fluctuations that, while not going up to levels like 200 mg/dl, are much less predictable than many people are told they should be.

Blood glucose levels are influenced by so many factors (Elliott & Elliott, 2009) that I would be surprised if they were as smooth as those in graphs that are frequently used to show how blood glucose is supposed to vary in healthy individuals. Often we see a flat line up until the time of a meal, when the line curves up rapidly and then goes down quickly. It usually peaks at around 140 mg/dl, dropping well below 120 mg/dl after 2 hours.

Those smooth graphs are usually obtained through algorithms that have statistical methods at their core. The algorithms are designed to generate a smooth representations of scattered or disorganized data points. A little bit like the algorithms in software tools that plot best-fit regression curves passing through scattered points (e.g.,

The picture below (click on it to enlarge) is from a 2006 symposium presentation by Prof. J.S. Christiansen, who is a widely cited diabetes researcher. The whole presentation is available from: It shows the blood glucose variations of 21 young and normal individuals, based on data collected over a period of 2 days. Each individual is represented by a different color. The points on each curve are actually averages of two blood glucose measurements; the original measurements themselves vary even more chaotically.

As you can see from the picture above, each individual has a unique set of responses to main meals, which are represented by the three main blood glucose peaks. Overall, blood glucose levels vary from about 50 to 170 mg/dl, and in several cases remain above 120 mg/dl after 2 hours since a large meal. They vary somewhat chaotically during the night as well, often getting up to around 110 mg/dl.

And these are only 21 individuals, not 100 or 1000. Again, these individuals were all normal (i.e., normoglycemic, in medical research parlance), with an average glycated hemoglobin (HbA1c) of 5 percent, and a range of variation of HbA1c of 4.3 to 5.4 percent.

We can safely assume that these individuals were not on a low carbohydrate diet. The spikes in blood glucose after meals suggest that they were eating foods loaded with refined carbohydrates and/or sugars, particularly for breakfast. So, we can also safely assume that they were somewhat "desensitized" (in terms of glucose response) to those types of foods. Someone who had been on a low carbohydrate diet for a while, and who would thus be more sensitive, would have had even wilder blood glucose variations in response to the same meals.

Many people measure their glucose levels throughout the day with portable glucometers, and quite a few are likely to self-diagnose as pre-diabetics when they see something that they think is a “red flag”. Examples are a blood glucose level peaking at 165 mg/dl, or remaining above 120 mg/dl after 2 hours passed since a meal. Another example is a level of 110 mg/dl when they wake up very early to go to work, after several hours of fasting.

As you can see from the picture above, these “red flag” events do occur in young normoglycemic individuals.

If seeing “red flags” helps people remove refined carbohydrates and sugars from their diet, then fine.

But it may also cause them unnecessary chronic stress, and stress can kill.


Elliott, W.H., & Elliott, D.C. (2009). Biochemistry and molecular biology. 4th Edition. New York: NY: Oxford University Press.


Anonymous said...


Good post. It would be interesting to see what the serum blood glucose response would look like during a normal "American" commute to or from work. I bet catecholamine response to the stress of the commute would cause some significant fluctuations in blood glucose. It makes my wonder about the potentially damaging effects of chronic fight/flight response without significant physical activity component to go along with this hormonal response. I am starting to think I should rate the invention of the automobile right up there with agriculture as one of the biggest mistakes made by human beings. Then again, maybe I am being a little harsh. What do you think? I wish I had the opportunity to walk or ride a bike to work everyday :( Thanks again for another thought provoking post. BTW, I still need to get that glucose meter so I can do some experiments of my own.


Chris Kresser said...


The question we should be asking isn't whether these people are diagnosed with diabetes, both of which are based on arbitrary and unjustifiably high cut-off points. The more relevant question is what blood sugar levels have been shown to cause health consequences and complications normally associated with diabetes.

Jenny Ruhl has written about this extensively on her site, Diabetes 101. This article explains why the diagnostic criteria for diabetes are basically useless for preventing complications due to high blood sugar. While it may be "common" for blood sugar to rise above 150 mg/dl even in healthy subjects, that doesn't mean it's normal or healthy.

In fact we know from studies that blood sugar >150 mg/dl an hour after eating strongly predicts heart attack. We also know that there's a very tight correlation between heart attack risk and A1c levels at as low as 4.6%. See here for more info.

So it seems to me that while these fluctuations you describe may be common, elevated blood sugar readings above 140 mg/dl an hour after meals are certainly something to avoid.

Ned Kock said...


I think you are right about car commuting in general. I would also place in that category stressful online interactions (e.g., flaming).

We experience too much stress while sitting down. I would recommend at least standing up and walking a bit, after a stressful situation.

Indeed, a genetically programmed stress-fight/flight response, without anything like fight/flight cannot be too good for us.

By the way, cortisol levels seem to get much more elevated in response to stress after a hyperglycemic meal than after a meal rich in protein or fat. This needs a separate post, but it makes sense to avoid refined carbs or sugars before that commute.


Some of those folks who experience ketonuria (ketones in the urine) in the morning, also have high fasting glucose levels. The cause is physiological (or benign) insulin resistance.

Physiological insulin resistance is also common after heavy exercise. My goal with the post is to show that the numbers vary a lot, so there is no need to panic due to certain readings.

Having said that, you (and Jenny) are absolutely right about the problems associated with elevated blood glucose levels. Those levels are also associated with decreased lifespan:

generic viagra said...
This comment has been removed by a blog administrator.
generic viagra said...
This comment has been removed by a blog administrator.
viagra for sale said...
This comment has been removed by a blog administrator.
viagra substitute said...
This comment has been removed by a blog administrator.
Ned Kock said...

Four spam comments above deleted.

diabetes blue foot said...

Thank you for this information, I believe just as your statistics show that blood glucose levels vary greatly among healthy people.

The ADA says up to 100 can be consider normal fasting blood sugar where as other websites will say it should not be above 90 fasting.

There is no cookie cutter approach to what is "normal" blood sugar levels, this is the same problem we have with BMI being used as a measure of obesity.

Correlation does not equal causation and this why only a holistic assessment of an individuals health can tell us if they are healthy. Many who are called "diabetic" today are probably not. There is a diabetes epidemic in the West
but we must be careful with assuming 100mg/dl fasting is in fact Pre diabetes.

Anonymous said...

It's good that people are recognizing that there isn't a single range that is "normal" for blood sugar. I have non-diabetic Reactive Hypoglycemia can my normal range is 145-180. If I get below 140, I know I'm getting too low. If I followed the accepted "normal" range, by the time I got my sugar down to where it was supposed to be, I would be passed out. Other factors need to be taken into account before determining what your "normal numbers" are. I have an extremely high metabolism; I eat around 5,000 calories a day. Each individual is different.

Unknown said...
This comment has been removed by a blog administrator.
angelapostea said...
This comment has been removed by a blog administrator.
Frank Neill said...

Hi Ned
I’ve been trying to get figures for blood sugar distribution in the “normal” population (i.e. those without diabetes or pre-diabetes). I realise that the figures are up and down and all over the place, as in the graph in your article. What I want to know what percentage of people have average blood sugars in the 51-60 range, the 61-70 range, the 71-80 range, the 81-90 range, the 91-100 range. Ideally, I would like this not just for our “European” cultures, but also for those still following their “pre-MacDonalds” eating patterns (eg, people from Japan, maybe even Okinawa).
I am aware of the fact from my experience that some people seem to have blood sugars that tend to settle into the lower range between 1-3 hours after eating (and often depending on what they ate, too). Others, however, seem to settle into the mid or higher range of people who don't have blood sugar levels in the pre-diabetes or diabetes ranges.
I also realise that there may well be a variety of factors that determine how blood sugars move for individuals (including genetics, epigenetics [and this may be a significant factor, I suspect], pre-birth nutrition, early nutrition, what's happening endrchronoligically, and other facts as well).
I'm still keen to find out any trends that are known.
Also, I remember reading somewhere (possibly Purnell's History of the Second World War), that Japanese soldiers were still operating on blood sugar levels where European soldiers would have been comatose. If true, this is interesting.

Gretchen said...

Hi Ned,

I agree that there's a lot of individual variation in BG levels. But I've always wondered if the people with the very high postprandial levels in Christiansen's study went on to develop diabetes with age.

Today, more and more people who can afford it, even those without diabetes, are using continuous glucose monitors. It would be interesting to pool the results from the CGMs as Christiansen did.

Anonymous said...

"So, we can also safely assume that they were somewhat "desensitized" (in terms of glucose response) to those types of foods. Someone who had been on a low carbohydrate diet for a while, and who would thus be more sensitive, would have had even wilder blood glucose variations in response to the same meals."

why does this happen?

Should low carbers take actions to mitigate the wilder glucose response? If so, what should they do?

thanks for re-posting this.

Tucker Goodrich said...

The title of the chart says two days, but the times indicated are only for one.

Are both days overlaid?

Anonymous said...

Was the insulin response in these folks measured? Just because these BG responses are common does not mean they are normal or healthy. ON the Jocko podcast Dr. Peter Attia recently discussed the goal to minimize BG spikes and that the true goal was to minimize insulin area under the curve. Accepting the premise that these spikes are harmless seems unsupported and may well mean more are close to ill health than realized. For me intermittent fasting and weight lifting have improved blood glucose control. I'm Thin and low carb but formerly desk bound for 20 years. I Don't take comfort from this analysis but will continue trying more fasting and lifting to see better BG numbers. Colleen

Ned Kock said...

Normally the areas under the curve (AUCs) are proportional to the amplitude of the spikes, but more so to the HbA1c’s. The HbA1c’s in this study varied from 4.3 to 5.4 percent.

Anonymous said...

Remember the "glycemic index"? (a dietary approach that ranked foods according to the glucose spikes they produced (maybe the area under the curve rather than the spikes).
In view of the wide variations of BG response to food input, I wonder if the glycemic index makes any sense: you could give exactly the same food to two individuals and their BG response could be very different, destroying the whole idea of ranking foods by their glycemic index.

Gretchen said...

I haven't researched this lately, but my understanding was that yes, different people would have different BG responses to the same food, but the *relative* responses would be similar, with some exceptions. In other words, A might go up 10 points with food X and 20 points with food Y. But B would go up 25 points with food X and 50 points with food Y.

BTW, anyone who got too much of people named A, B, and C, would enjoy Stephen Leacock's piece in Literary Lapses: I giggle whenever I reread it.