HbA1c: do the latest BG samples count for more?



As for recent weeks having more impact, I think it is the opposite, according to a paper that came out a couple months ago. The first 30 days of your last 120-day period should account for about 50% of your result. And the last 30 days of the 120 should only account for about 5% of the result.
If you have a source that says different, please let me know.

How Glycemic Variability impacts HbA1c

I would be surprised to find a study stating this conclusion:

  • the HbA1c measures glycated hemoglobin, which is found in red blood cells.

  • red blood cells die out over an 8-12 week cycle, using a probability curve for the die-out (likely Gaussian, although I don’t have proof on that)

  • a sample of red blood cells for the A1c will therefore count more newer cells than older cells

  • therefore, more recent times in the 90-day window will be overemphasized when measuring an A1c (or anything based on red blood cells).

  • This is also the understanding presented by every endocrinologist I have ever talked to.

I would need to read a very well researched and documented study if it were to present such a counterintuitive result before believing it :slight_smile: I am not even sure that one study would be enough.

Meta Study: Long-term Glycemic Variability and Risk of Adverse Outcomes
How Glycemic Variability impacts HbA1c

This is my understanding as well, @Michel – I don’t have time right now to dig up the relevant studies but maybe I can look at home.


Yes, please provide them.
And thank you for putting the work into splitting the topics up.

It seems intuitive when considering the red blood cells as dying at a consistent rate, but that ignores how the glycation bio-accumulates (oldest RBCs will be most glycated). It was a paper to predict future, or “real” a1c values if testing under 120 days.

There is a lot that seems intuitive or true that I have discovered lately to be myths (Metformin’s unknown (or lack of) effect on peripheral tissues, Diabetics having too much glucagon production instead of too little, a NASA study saying ideal BG for cognition is around 120-150mg…)

The paper uses some complex math, and is not free, so I was going to try to print and mark it up before sharing. But here it is if you can manage access:

Use of a model for A1c formation to estimate average glucose operative between short-interval measurements of %A1c.
Clinical Biochemistry [Volume 54, April 2018, Pages 73-77]

Here is a classic, old article in Hba1c kinetics that says the first 2 months of the period are most important:

And if you look at Figure 5 in this article, you’ll also see how the first 60 days of the 120+ day period (some RBCs last longer than you think!) account for most of the change.

I’m going to need to cut back my time on internet forums for the next week or two while I work on some papers and organizing. So don’t think I suddenly disappeared!
Happy Reading!


Michel, you are correct. The most recent BG has a greater impact on A1C.

Indeed, dead RBC’s have no affect on the A1C reading, as they are removed from the body. And certainly a greater number of RBC’s from 2 months ago will be dead than the newer RBC’s.

This link clarifies it:

Will the A1C test show short-term changes in blood glucose levels?

Large changes in your blood glucose levels over the past month will show up in your A1C test result, but the A1C test doesn’t show sudden, temporary increases or decreases in blood glucose levels. Even though A1C results represent a long-term average, blood glucose levels within the past 30 days have a greater effect on the A1C reading than those in previous months.


Ok, I figured it out. Reviewed all the relevant research on Hb glycation kinetics. We are talking about the same thing from opposite directions.

Some of the more informal articles are a little confused talking about how there are more young RBCs than old ones (wrong). Assuming a 120-day lifespan (this isn’t 100% consistent), then your past 30 days of young RBCs are only 25% of the total, and their glycation is much less, since they haven’t been around long enough to be glycated much!

IF you have a relatively steady glucose for the next 120+ days, then the next 30 days will predict 50% of the result. If you had an average glucose of 300mg for the last 120 days (t-120 to t-1), but suddenly, tomorrow (t+0), you get it controlled to 150mg, and keep it at that level for 120 days (t+120), then the next 30 days will account for 50% of your final result. You’ll have much more old cells than new cells (t+120) and all old cells will be much more glycated over time, so they predominate in number and strength.

So, if you are talking about a big change today (t-0), then your result after 30 days will have a big change, because 30 days worth of your oldest RBCs (very very glycated, and largest factor in your t-0 result) will have died. Replaced by 30 days of new, fresh, relatively unglycated RBCs.

I hope this makes sense… It is complicated. No time to link and describe all 10-15 papers, but here are some more images that I hope will make this more clear.

The paper most of those sites (and some papers) link to and get confused about:

How a big change in BG control on day 0 affects the next 120 days:

The rate at which RBCs accumulate glycation over 120-day lifespan. Top is a steady average 100mg (5.1% a1c), bottom is 300mg (12.1%). Note how the 120-day old RBC have an a1c of over 10% (for 100mg avg!) and 30%, respectively!

The rate for different non-diabetics (NDM) and diabetics (DM).

If this is still confusing, imagine what Day 119 would look like and graph it out on some paper or spreadsheet. Imagine every day you get 1 new RBC, and the ones 119 days-old have 3x the glycation of 30 day-old RBCs.


I think the best way to know for sure is to take the standard glycation equation model, use a recognized estimate for the parameters, kg and gHb(0)/tHb, then to run a simulation.

Here is a recent article (2016) that describes glycation kinetics:

The estimated parameters (listed in the Supplemental Methods) in this article are:
E{kg} = 6.07 x 10-6 dl/mg/day
E{tHb(0)/gHb} = 0.0032 (very small)

So anyone is welcome to calculate it and present their results. I intend to do so myself in the next couple of days if I find the time.


A side note is that you also need to know MRBC, the average 1/2 life of the red blood cells. The same article above estimates it mathematically in the Supplemental Methods for the specific study at E{MRBC} = 53 days, giving an average life for these patients of 106 days for their red blood cells.

The same Supplemental Methods mention:

[…] prior methods of estimated glycation rate constants […] often neglect rate cell aging or assume a fixed MRBC (of 60 days ) for all RBCs and all patients […]

This means that many prior studies of glycation assumed RBCs to live 120 days on average. But we knew that, as exemplified, for instance by:

[…] in real life RBCs live about 120 days […].

or by

The maximal lifespan of erythrocytes is 120 days, with a rather small variation of approximately 10 percent.

So we should be able to put to bed the part of the discussion that discusses average RBC lifespan.

Note, however, that very same article uses custom estimates PER PATIENT for RBC lifespan to evaluate Average Glucose from HbA1c, and concludes that it explains all variations that bother us all so much (thanks to @Kaelan, btw, for discovering this article). But, based on their results, I think it is reasonable to use either 106 days or 120 days for a reasonable approximation of a simulation (or really anything around 120 days…).

Of course, if we compare several simulations, all models, to be comparable, would have to be recentered on the same number for RBC lifespan, for which I think 120 days is a good value.


Seems like a lot of extra work for a 5-15% variance from a simpler estimate, but if anyone wants to take a crack at it, we’d be very interested in the results.


You can get an estimate of RBC lifespan by doing a reticulocyte count, and from your hematocrit.

Usually you get hematocrit whenever they do a complete blood panel, so the only thing “extra” you would need to ask for is a reticulocyte count.

You lookup your hematocrit value in the corrected reticulocyte count table, and that gives you your reticulocyte life span.

Then you plug in this formula:
RBC survival in days = 100 / (reticulocyte count (in percent) / reticulocyte life span)

Again, this is only an estimate, but it is certainly closer than just using the “average” for everyone!


@oni, I just read in detail the references you gave in your justification of your statement that HbA1c is backwards weighted.

  1. Your first paper I could not get to, as it is, as you indicated, behind a paywall.

  2. Your diabetes Care 1995 article, of which you write:

But the article actually says the exact opposite of your statement:

The weight functions for glycated proteins had maximum values on the days just before the measurement of glycated proteins and gradually decreased with an increasing time interval. […]
The lengths of the periods over which the weight function[…] for HbA1c [… was] estimated to be roughly 100 days […].
CONCLUSIONS The level […] of HbA1c [does] not reflect the simple mean but reflect the weighted mean of the preceding plasma glucose level over a considerably longer period than was previously speculated.

Translation: what this means is that the HbA1c is a weighted average of the past blood glucose levels, and the weight of the most recent samples is highest (“weight functions for glycated proteins had maximum values on the days just before the measurement”).

  1. Your PMC article, of which you wrote:

But, when I read through the paper, again it actually expresses the exact opposite of what you wrote. In the quote below, the authors discuss Figures 5A-C, with profiles having the SAME 120-day average glycemia:

Based on these figures it can be observed, for example, that 60 days after a drop of BG from 22.2 mmol/l to 5.6 mmol/l, HbA1c is equal to 7.9% (62 mmol/mol), whereas 60 days after a rise in BG from 5.6 mmol/l to 22.2 mmol/l, HbA1c is equal to 12.5% (113 mmol/mol), despite the same average 120-day glycemia in both cases.

What this means:

  • in case 1, the first 60 days were high, the last 60 days were low, HbA1c = 7.9%
  • in case 2, the first 60 days were low, the last 60 days were high, HbA1c = 12.5%
    Conclusion: the last 60 days were much more important than the first 60 in both cases for HbA1c, despite the same average glycemia in both cases

On that basis, I don’t think we need to do any simulation, since all the references say the same thing: it is the later part of the last 120 days that is preponderant in the measurement of average glucose by the HbA1c.

But I am disappointed that your interpretations of the quotes you gave us were the opposite of what they really mean :frowning: Readers cannot always review every single reference given, and need to be able to trust someone’s interpretation of them. This makes me doubt the meaning of the other references you gave in other current threads, especially on the heels of this other instance where the reference you gave was also different from the interpretation you gave of it:

So I think it is really important to read through a study and understand it thoroughly before quoting it for a purpose! Not that I have not been caught by that before…

As for the point of this thread, I want to make sure to repeat the conclusion here for any stray reader:

The HbA1c is biased towards the latest part of the last 120 days of average glucose levels. The last 60 days count for more than the first. The last 30 count for more than the next 30, etc.


I disagree with your interpretation, actually. It’s confusing how they word it, I know. Caught me off guard in the beginning too.
Again, we may be saying something very similar, but from opposite directions.

I think I’ve explained how it works thoroughly. And the logic of it: the accumulation of glycation, and the preponderance of older euthyrocites than newer (about 75% more 30-120 day old euthryrocites than 0-30 day relatively unglycated ones).

The paper you cannot access also explains the math in more detail. It specifically estimates A1c of 120 days in the future using the next 30 days, and graphs the accuracy.

But others can decide for themselves what makes more sense.


To put another way, what makes the last 30 days of euthryrocites important is not how important they are, but how important the death of euthrocytes from days 120 to 150 (and your blood sugar then) were.

Your euthryrocites born yesterday have almost no Hb glycation yet, the ones from 119 days ago have thousands of times more glycation, but yesterday’s will still replace them tomorrow, making their effect progressively 0, no matter how weak an effect the babies had in comparison.


The text of both articles you posted that are available for all to read is VERY clear, and I have quoted the relevant quotes word for word. There is no possibility of ambiguity, sorry.

I understand what you are saying. But what you are saying is NOT what your articles are writing. Again, I quoted word for word the relevant passages, and no ambiguity is possible. To reinforce it, here is one of the passages from your own article:

The weight functions for glycated proteins had maximum values on the days just before the measurement of glycated proteins

There is absolutely no ambiguity: the weight of the last samples, just before the measurement, is highest, and counts for most. I could also re-quote your second article and show the same thing again, but why, since I already did it 3 posts above?

I am not trying to be rude, I promise—but your explanation has nothing to do with what your articles are writing. It is neither here nor there, and contradicts your own articles.

Science is not like Comparative Literature. There are, in general, no two ways to read an equation: this is what we are talking about here. Your two articles and @Eric’s two articles all say the same thing: it is the later part of the 3-month measurement period that counts for most. It is impossible to read 4 articles that all say the same thing, and then argue that they are saying the opposite :slight_smile:

So please, before arguing again against the very conclusion of your articles, have a good look and read them again. In particular, read the sections I quoted: it is impossible for anyone accustomed to reading scientific literature to make a mistake in this case. These sections make it crystal clear :slight_smile: Let’s remember that thousands of PWDs read these threads. We cannot lightly spread misinformation: we can cause serious consequences if we cause people to mistreat because of what we post.

[EDIT] I am complaining about these articles, but I really like the fact that you are sharing a bunch of them though! Data = better treatment for all of us :slight_smile:


@Eric, very cool, I had never realized that!


I posted about it months ago, but I think you were in a non-internet area of the world at the time! :grinning:

If you search FUD for posts by Eric that contain “reticulocyte” it should be easy to find! I don’t use that word too much!


Wait, is that your interpretation of figure 4 in that paper? You read it as a timeline of the last 120 days?

It does use math, but from looking briefly at it I doubt whether it qualifies as complex math.


For those who are interested, but don’t have access to the paper: Semantic Scholar has the images. Figure 4 is the relevant image here. On the x-axis we have days since the change in average glucose and on the y-axis the ratio of observed change in HbA1c to the expected change in HbA1c after 120 days, provided that the conditions remain constant. As you can see, 30 days after the change in average BG, you have reached 50% of the expected difference in A1c after 120 days. Thus I have to disagree with the idea that the last 30 days only account for 5% of your A1c.


@Boerenkool, to make sure I am getting it right, since we don’t have access to the whole paper but just the summary and the graphs, is this what you understand for the first point, for instance:

  • if you make a change of BG, then wait 30 days to an A1c, the (short term) A1c you get already reflects 50% (roughly) of what the final A1c for the new BG would be, 120 days after the change?

That is what I understood and that is also what I read in your comment.

In that case, what this would say is that, if you have a change of (average) BG 30 days before the end of your A1c interval, your actual A1c after these 30 days will be reflective of 50% of the change to what it would be if you stayed at that level for another 90 days, then measured the A1c.

This is largely equivalent to saying that the last 30 days of an A1c measurement are about 50% weighted (instead of 25%)—again, saying the same thing as the other 4 articles mentioned in this thread, but this time with a quantitative analysis.


Exactly, if a step change of average BG is introduced at t=0 and kept constant, then the observed difference in A1c at t=30 reflects 50% of the predicted difference in A1c at t=120.

Right, in that case your A1c at the end of your A1c interval will be reflective of 50% of the change in A1c between 30 days before the end of the interval and 90 days after the end.

Higher average eAG, lower A1C?