Inaccurate A1C Measurements

Has anyone ever received an A1C value they just don’t trust?

My A1C has been hovering between 5.4-5.9 for about three years now. My most recent value from last year was 5.5. I just had it tested yesterday and expected it to have increased slightly because, well, you know, I can see what’s happening on my Dexcom and I just haven’t had great control lately. But shockingly the test yesterday came back at 4.8. While I understand that would be an amazing value if true, I just don’t believe it’s accurate, especially because my Dexcom says my average value over the past 90 days is 124.

All my tests have been at a lab, I’ve never used an at-home test kit. It’s a different lab this time than last time, but still, one would expect consistency. Has anyone ever experienced issues with inaccurate lab values?


Different labs may give different results. Try to return to your
original lab next time.

Online estimators have 124 average glucose as 5.9% A1c

I don’t have a CGM but…is it possible that it’s the Dexcom that’s inaccurate?

Yes, Dexcom is inaccurate when A1C is in that low range. President Sayer of Dexcom said at his last earnings call, that in the future they do plan to work on a CGM that will be accurate for patients that mostly keep their BG between 70 and 120. A1C also varies between labs based on the equipment they use and certifications ±.5% so this could have just been a difference between labs as well or a combination of both variables.

Since the A1c number is expressed in one tenth of a percent resolution, it seems to imply a precision that is just not there. I read in one medical paper that an A1c of 6.0% actually corresponds to a relatively wide range of average glucose. If memory serves me, it ranged from 100-152 mg/dL with the median at 126 mg/dL.

Your 4.8% A1c is likely an outlier given recent measurements in the 5.4-5.9 range, but measuring again in 90 days can help you put this reading in perspective. Many things can interfere with the accuracy of an A1c reading, including shorter or longer red blood cell life or iron deficiency anemia, to name a few.

My dexcom estimated I would be at 6.3 the last test and it was 5.8. That’s a little lower for me too
However my a1 c is always about .5 lower than that estimate suggests.
However a 4.8 is difficult to achieve unless you are experiencing lots of lows.
Some people seem to do it tho

I agree with @Terry4, it sounds like you have an outlier data point.

Here is a reference to a frequently cited paper on this topic:

Many people (and many online eAG to HbA1C calculators) remember precisely two sentences from this paper:

The following formula describes this relationship: 28.7 x HbA1c ‐ 46.7 = eAG. This formula will assist health care providers and their patients in interpreting HbA1c values in units similar to those that the patients see regularly through self‐monitoring.

Ooh, 28.7 and 46.7! It MUST be accurate you think.

Not so fast!!! Here is the actual data from that paper from which that equation was extracted:


This shows that you can fit a straight line though ANY set of data and extract seemingly precise numbers for slope and intercept … regardless of the amount of scatter in the data.

Based on my several years of CGM experience, I know that my CGM AG will predict a HbA1C that, this far, has always been 0.6-0.7 LOWER than my lab HbA1C will measure. As noted elsewhere in this discussion, others see an actual HbA1C that is LOWER than their CGM AG would predict based on the “magic” equation.

But, I still agree with @Terry4 … it seems as if you most recent result was “funny” for some reason.

Stay safe!


1 Like

Go ahead and have it retested if you are worried about it. You could see what the at home A1c reader says, but lab is better than anything else (including Dex).

I have called the lab before to find out what equipment they use and get the associated accuracy. You can do that. I think its unusual, but they will tell you.

Thanks for all the info. I’m not sure if I’ll have it retested or not because at the end of the day it doesn’t really matter what that value is. It’s just one of many numbers associated with how I manage my T1D. I’m generally healthy, which is the most important thing. I’ll have my A1c measured again in 3-6 months and see what’s changed.


To follow up on this I just had my A1c retested at a different lab and got a value of 5.6, which is where I would expect it to be given my recent history. So one lab says 4.8, another says 5.6? That’s a pretty big discrepancy in values. I guess the takeaway from all this is to not use that lab again.

The correlation between BG and HbA1C used in AGP reports is different from the older equation that is used in calculators. The change in HbA1C between samples analyzed at the same lab has some importance, as it indicates changes in levels and controls. There is a LOT of biological diversity (see graph in Joh_S2 post). It is really time to move on from HbA1C to Time in Range.

Is it poor execution of a protocol that undermines your trust in this particular lab or could it be that protocol itself is flawed and that it implies a 0.1% precision that is just not there?

My perception is that 1/3 of A1c’s are lower than the glucose experience, with 1/3 higher and 1/3 fairly close. So, in diabetes forums, 2/3 of people’s experience with A1c is that it closely correlates with their average BG or gives them a better A1c than expected.

I’m in the 1/3 whose A1c is higher than the average glucose justifies. I’m not happy with this number being used as shorthand between patients, doctors and academics to characterize glycemia. Any other situation where 2/3 of the data was bad (1/3 lower + 1/3 higher) would not long be used.

As people with diabetes I think we would all be more accurately served comparing glycemia denominated by TIR and glucose variability when CGM data is available.

1 Like

I totally agree and when we are using the same hardware and software, like the Dexcom G6, we are so to speak using the same lab (Dexcom) to compare our results. It just takes one or more of the variables out of the loop.

1 Like