Which Is Better…a Lab A1c Test or the Estimated A1c (GMI) from a CGM?

Dear Dr. Edelman,

My most recent lab A1c test was 7.6, but my 90-day CGM report (GMI) said my estimated A1c was 6.9. What’s the difference between the two, how can they be off by so much, and which one should I trust?

Dr. Edelman:

The A1c is the traditional laboratory test to determine what your average blood sugar is over the past 2-3 months. It has been known for quite some time that this lab test is commonly very inaccurate, and can be affected by many different situations such as pregnancy, ethnic background, kidney disease, liver disease, certain blood dyscrasias, etc. In fact, the HbA1c test can be off by .7-1%! On the other hand, the glucose management indicator or GMI (estimated A1c) is a number that represents the average blood sugar over the past three months because it is derived from the actual blood sugar from a CGM. My new favorite phrase is: “The best way to determine the average blood sugar over the past 2-3 months is to measure the average blood sugar over the past 2-3 months.” Which is exactly what these CGM devices do. Lastly, I believe the lab measurement of the A1c should be called an “estimated” value and the GMI the true value. In reality, for people who have a CGM, there is no need to get your blood drawn for the A1c except for insurance purposes, because they’re still living in the 1920s.


Additional Resources:

Why the A1c Sucks (And Why Time in Range Is More Important)

  1. Avatar

    This is quite interesting. I am using the Libre2 system and for almost all sensors the Libre values are lower than blood glucose determined by finger sticks. I have a friend who has made the same observation, and there are comments alll over he internet that the Libre sys “reads” lower than finger sticks on a consistent basis.

    In my case the Libre sysem recoods an average of 10 points lower than the corresponding glucometer values. In fact, I “calibrate” libre sensors vs. glucometer values for each sensor so that I know how much different they are on average.

    The main “probllem with this is that I often alarm as low on the Libre sensor when in fact I’m usually a minimum of to points higher.

    So it’s difficult to know what to “believe” – A1C directly measured vvs. A1c estimated from CGM vs. A1c estimated from glucometer reads (back-calculating from the “calibration” diffeence betweeen CGM and glucometer).

    • A lot of people do comment on that – it’s been known that the Libre does read lower than the actual value, and that could mess up your GMI. For you the lab A1c method may be more important.

      • Hi Dr. Edelman – I just went to get donuts after watching your video today. 🙂

        I am newly diagnosed with CKD and my labs show my A1c to be between .5 and 1 higher than my GMI of 6.6. My TIR is about 84%. Can I still rely on my GMI for better accuracy, or does my decreased kidney function actually affect my glucose-hemoglobin making my labs the better indicator?

        I have both an endocrinologist and nephrologist, but won’t be seeing them for a few weeks, so just curious about your thoughts. Thanks.

        • It’s actually the opposite – having kidney disease, depending on how severe it is, can make your labs inaccurate. For people with any medical condition that can affect the labs, you definitely want to use the GMI because that’s not affected by kidney disease.

  2. Avatar

    I have not been able to figure out how to download information from my Dexcom to give to my primary care people. Or is that not possible? Also, where do I find the GMI (glucose management indicator)? This is my second year with the Dexcom and into almost a year with the Dexcom G6. It changed after my first full year with the Dexcom.

    • Avatar

      Hi Marilyn,

      I’m NOT associated with Dexcom except as a user. But I’ve had T1 for almost a half century, and have been using a Dexcom CGM since their first water-resistant model (thanks to Steve). I’ve been really impressed by its improvements in accuracy, miniaturization, and ease of use over the years. I’m not as impressed with the fact that as they have emerged as the 600 pound gorilla in the world of CGMs, their emphasis has (somewhat) shifted from the users to the stockholders. But they haven’t abandoned us.

      Since you’re obviously on the internet, my first suggestion would be to go to “https://www.dexcom.com/guides” on the Dexcom website. In particular, I’d suggest looking at the links and downloads for both the G6 and Clarity subjects. You might just find the answers you’re looking for from that webpage. They also have a number of videos on YouTube about how to use their products.

      However there are a number of issues that might affect the answers to your questions:
      Do you have a compatible “smart (aka: mobile) device” like a cell phone?
      If not, are you using the dedicated Dexcom Receiver?
      Do you consider yourself “internet savvy” (i.e. comfortable with downloading and using various applications on your computer and/or mobile device)?
      Do you have an endocrinologist or at least a CDE (Certified Diabetes Educator – possibly an RN or other diabetes specialist), or do you receive your diabetes-related health care from your primary care provider.
      Does your health care team have and know how to use the Dexcom Clarity app (the doctor’s version).

      I believe the majority of G6 users use a mobile/smart device (usually a cell phone), instead of the dedicated Dexcom receiver. But to use your phone, you need to have downloaded and installed the G6 application, and then paired your Dexcom transmitter and current sensor with it. The sensor data is transmitted to your cell phone, where you can see it in real time on the G6 app, but only up to the last 24 hours. The raw data may continue to be stored on your phone by 3rd party apps if you want, but it is intermittently transmitted up to Dexcom’s secure data servers (via “the cloud”) when your phone is connected to the internet.

      For the most part, the G6 app on your phone provides only a short-term, real-time view of your BG data. To look at a greater range (i.e. weeks, months, or years worth) of data, you need to download the Dexcom Clarity app. You can do this on your cell phone and/or a laptop or desktop computer. The Clarity app then connects you with all of your data stored “on the cloud”, and provides a wide variety of ways to segment, filter, display and download your BG values as reports, along with any other data you input (like carb counts, insulin dosages, and exercise).

      The more data you add, the better will be the overall view of your diabetes management efforts. If your doctor’s office has the Dexcom MD version of the Clarity app on their computer(s), you can generate an authorization code to allow them to access your CGM data directly from the Dexcom servers. As long as your authorization is in effect, they can access the data at any time. Thus you won’t have to download your phone or do anything else to give them your BG data. You can also revoke your authorization if desired.

      If your doctor doesn’t have the Clarity app at their end, you’ll still need to use Clarity at your end to view (and download) the data in your desired date ranges and output report formats (graphs, lists, etc.). When you view those Clarity records (online) you’ll have the option to download and save them as PDF reports on your phone or other computer. You can then use the free Adobe Reader app (or others) to open and view those reports any time you want.

      You can then ask the doctor’s office what date ranges and report formats they would prefer to review, and send them as e-mail attachments before your appointment, or print them out to take with you and give to the doctor.

      Dexcom also has a team of educators that you can contact via a link on their website. I’m pretty sure that their service is free of charge. In some areas, I believe that they have reps who can work with you or a diabetes support group in person, but I’m sure there are also options for a phone consultation or e-mail exchange.

      If my answers weren’t helpful, Dexcom will always be your best source for the best answers.

      Good luck, and stick with it.

    • Avatar

      Look for the Dexcom Clarity app. That will give you all of the information.

  3. Avatar

    This Pandemic has really hampered many of the things that had gotten use to doing. I really miss attending the TCYOD Meeting in Washington, DC whenever it rotated around to us. I still have many of the materials I obtained and purchased during my attendance. I often think about the Happy Diabetic. I was asked by my church twice to make a diabetes presentation and I credit much of my early learning as a diabetic from the TCYOD Sessions. Looking forward to whenever you return to the DC area.
    I enjoyed your competition.
    By the way, I am a type 2 diabetic and so far I am doing okay with my management. My A1c is one or two points around 7.
    Take care and stay safe.

  4. Avatar

    What do u do when your UCLA endo and Cedars Sinai endo demands a lab A1c? And states the CGM is not accurate?

    • It’s time to stop bitching about your UCLA endo and switch! Unfortunately, insurance companies still require a lab test of A1c, but there’s no question the GMI is way more accurate than the lab.

  5. Avatar

    I’m a T1D medical student (MS3) and I’m PI of a study where we looked at 90 day readings for about 60 people with T1D and their self reported most recent A1C was remarkably similar to the estimated A1C (although estimated A1C was slightly higher). I was surprised by this because my actual A1C is always much lower than the estimated A1C by Dexcom, I partially blame that on my Gilbert’s since there is a negative association with bilirubin and A1C. I think it’s worth considering that the estimated A1C takes into consideration the outliers (post prandial spikes) when calculating the estimated A1C, these outliers might be over represented in the mean BG and estimated A1C especially in a generally well controlled T1D that has post prandial spikes that return to euglycemia. We know that Dexcom’s (our study CGM) BG readings are not as accurate below 80, I speculate this could potentially reflect an inability of Dexcom to detect a falling BG from a post prandial high as opposed to detecting a rising BG more accurately. If this is true it would erroneously inflate the estimated A1C. It does seem more logical to hold higher consideration to an estimated A1C based off of ~25,0000 BG readings than an the actual A1C but until more studies are published about complication risks, ect, on estimated A1C I don’t know if other endo’s will follow suite. I think the mathematical complexities of this estimated A1C are worth considering.

    • Thanks for your comment. The most important thing with the laboratory A1c is that it can be extremely off (i.e. .8% above and .8% below the actual value) and there are so many different laboratory methods, you can’t make sense of the lab A1c not matching the 90 day GMI. It sounds like you’ve thought of a lot of good theories, but there’s no question that the GMI over the prior 90 days is extremely accurate compared to any lab method. Here is an article that was sent to me by the head of lab services, you might find interesting: https://pubmed.ncbi.nlm.nih.gov/18540046/

  6. Avatar

    I have had a low A1c but a much higher GMI and my Dr. thinks this is due to my red blood cells not lasting as long and I am generating new ones much quicker? If this is true maybe a way our bodies protect ourselves from high blood sugar? Do you agree and have you come across this in other patients?

  7. Avatar

    9 months ago, my kidney filtering was 34%. 4 months ago it was 59% because i stopped metformin and reduced fenofibrate to 45 MG a day from 145MG. I want to try to get it to 75% but will accept 60%. In 90 days, my sugars have never been above 150

    I drink 32 ounces of water a day and also diet soda likely 60 ounces a day.

    My last a1c yesterday was 6.7%. Does kidney function track on the A1C or is it a separate measurement. I do not eat a lot of carbs but will have an ice cream when i get hypoglycemic.

    Based on the A1C of 6.7%, is it possible to estimate kidney filtering?

    Is there anything I can do to increase kidney filtering beyond just drinking water? I do not get a lot of exercise

    • First of all, that is an impressive increase. However, an improvement in your glucose levels probably does not explain all that improvement. Metformin and fenofibrate do not affect the kidneys adversely. The A1c does track with all microvascular complications including eyes, kidneys and nerves, but over the long-term (years).

      Regarding your second question, “Based on the A1C of 6.7%, is it possible to estimate kidney filtering?” the answer is no.

      There are several things you can do to increase kidney filtering (such as controlling your blood pressure, etc) and several things to avoid (such as dye studies, etc). Watch the video I did with Dr. Bethany Karl (she goes into much more detail) here:

  8. Avatar

    I went online and got a free A1C test from Quest Diagnostics. My search criteria was free A1C test.

  9. Avatar

    Thank you. This is very helpful.

  10. Avatar

    I’m almost 20 weeks pregnant. My Libre 2 shows an average glucose of 141 for the last three months, about a 6.6 A1C equivalent. I had my A1C taken yesterday at a lab. It was 5.8! Quite the discrepancy. For the record, my A1Cs are typically between 6.5 and 7.0.

  11. Avatar

    Freestyle Libre has been proven to show a lower result than blood drawn tests so if you are using FSL then I would still suggest to get your blood checked quarterly.

  12. Avatar

    Dexcom’s technology is good to see something constant and not to have to jab your finger as much. Nonetheless, it reads blood differently than meters and a1c’s, and is often off by .2-.3. I know my body, so when I want to get my real readings according to how I feel at the moment I use my meters. I then have to calibrate Dexcom twice to get it right until it wanders off again. Dexcom made me ill a few times early on before figuring out it needed backup.

    • How do you know the meter is correct over the CGM? Saying that, if testing with the finger stick makes you feel safer, then it’s important to do that. Do what you have to do.

  13. I am on the dexcom g6 and it shows my a1c value at 7.5 but today I was at my Dr’s office and the ran a finger stick a1c and it came back 5.4. My doctor seems to think it’s accurate an was super happy but I am skeptical of this. I have been dieting my sugars have been a lot better than they have ever been but I do still run high sometimes so I am wondering why the BIG discrepancy. I use to work for said Dr’s office and would use the A1c machine on myself and it would be different then my cgm by a point or 2 but never this much. Can you please explain this to me I am hoping for a 5.4 a1c thank you

    • The first thing is, you have to make sure it’s a 90-day download from your CGM. It has to be a 90-day average glucose estimated A1c. There’s no question that it’s way more accurate than the A1c finger stick or clinic lab test. You get 12 measurements an hour with the CGM, so you’re calculating your A1c from almost 26,000 measurements over 90 days, which is very accurate. I know we all want the lower number, but there’s no question you should follow the glucose on the GMI.

  14. Avatar

    Thank you so much for bringing this question up and for explaining it. I felt so defeated the my GMI showed 6.5 and then my A1C came back 7.0 I’ve been trying so hard to eat better. But I did go from 7.3 a1c to 7.0 in three months just changing my eating no medication change so I’ll take that as a win.

  15. Avatar

    Interesting, as I thought the HbA1c blood test was accurate and everything else was approximate; however since I started using a Libre* I largely ignored it because I couldn’t really do anything with the information. So what if my number from the previous 90 days was x or y? How does it really help?

    I know that in my early days I’d be given generic advice to adjust my basal, diet or exercise, but now I get real-time data on what’s happening (which I still find truly amazing).

    *Libre 2 with Bubble, so I get updates to my phone every 5 minutes, multiple alarms (and calibration).

    FYI: my LibreLink GMI currently shows 6.3% and the GlucoseDirect app I use with the Bubble (calibrated) shows 6.6% which is exactly the same as my last HbA1c taken two months ago (probably a coincidence).

  16. Avatar

    I had a HBA1c of 44 mmol/mol in March. I have been wearing a libre 2 to assess the way different foods spike my blood glucose. I have just had a repeat HBA1c which is now 32 mmol/mol 5.2%, my average glucose has been 5.7 over the last 3 months, but my GMI estimated HBA1 c is 5.8% 40 mmol/mol. A difference of 20%. I have a history of pre eclampsia x 2, and approx 25 yrs of hypertension medication mainly ACE inhibitors I am now 63. Does my elevated GMI result indicate kidney disease.

    • The GMI does not indicate anything at all regarding kidney function. You need a urine albumin to creatinine ratio, and an estimated GFR from your healthcare provider, which will tell you if your kidneys are in good shape or not. Having kidney disease will artificially lower the A1c, but will not change the GMI.

  17. Avatar

    I’ve been going through a similar roller coaster.

    I’m a 50 yr old Crossfit coach and endurance MTB racer. About 6 months ago I got an A1c result of 6.1. Since I don’t have a lot of room to improve weight or activity; I already train like a semi-pro athlete, I started using a CGMs from Nutrisense and Levels to figure it out. I also adjusted my nutrition from a carb heavy endurance athlete diet to relatively strict Paleo/Keto.

    While my CGM has given me an average blood glucose of appx. 97mg/dl for the last 6 months (corresponding to an A1c of appx. 5.0) A1c tests over that time (both home and lab) have come back between 5.8 and 6.1; indicating almost no change and estimating an ABS of appx 140mg/dl. This has been frustratingly confusing given how much I’ve adjusted in terms of nutrition.

    I’ve been wondering which numbers to trust. My nutrition, activity, BMI, and performance all appear to correspond with what my CGM is telling me, but my A1c is telling me something vastly different.

    This thread has been remarkably enlightening.

    • Glad to hear it!

    • Avatar

      I have a pretty identical story. A1c 5.8-6. CGM avg right around 100. Very fit, lean, lots of exercise, and barely eat anything “bad” that would raise glucose.
      Have you found what you trust?
      Does the A1c number matter?

Leave a Reply