Dexcom Clarity estimated A1C: how accurate?

Do you ever check the estimated A1C report on the Clarity app? If so, how accurate do you find it to be compared with actual A1C tests?

I will know more next week, but compared to a finegrstick A1c that I had recently, Dexcom Clarity is close, but estimated a little on the low side. (5.0 vs. 5.3)

mine was spot on when i got tested a couple weeks back.

My CGM average has always been a lot lower than my A1c. The Dexcom Clarity estimated A1c is based on the CGM average. In the past my CGM blood glucose average predicted an A1c in the 5.2%-5.5% range. My actual A1c ranged from 5.9%-6.1%. This puzzles me but I think that my red blood cells may live longer than the standard A1c assumes.

Keep in mind that Clarity only bases its estimate on the period used. The 14-day report only uses 14 days, for instance, and a blood draw A1c looks back over approximately the last three months.

I’ll get another blood draw A1c in a few weeks. This will be my first A1c since Clarity was released. I speculate that the Clarity number based on a weighted average of the 14-day, 30-day, and 90 day number will accurately predict the lab A1c for most people. I would weight the 14-day and 30-day Clarity numbers a bit more than the 90-day number.

Let me say this, I was shocked when my endo told me A1c and it was the same as Clarity. I showed him the report and he was also shocked.

I would be surprised if Clarity fails to predict most CGM users’ A1c within a tenth of a percent or so. Dexcom Clarity shared the DiabetesMIne Usability Innovation Award in the Design category with Tandem t:slim pump user interface. I’m thinking that it’s the only analysis software I need to help me steer my BG management efforts.

My most recent A1c based on drawn blood lab test (not fingerstick) was 5.3, while Dexcom Clarity reported an estimated A1c of 5.1 over 30 days. I was somewhat surprised how close this came out. Even setting measurement errors aside, given that A1c is just a proxy for weighted average bg over a period of time, an exact match should be considered a coincidence.

awesome A1c Dragan. Congrats!!

My latest A1C based using doctor’s office fingerstick was 5.2, while Dexcom Clarity reported an estimated A1C of 5.1 over 90 days. What was funny is that my Endo ask me what I thought my A1C was going to be and I told her Between 5.1 and 5.3 based on Clarity. She was shocked. I told her that if Medicare didn’t require me to visit her 4 times a year I could fire her. She laughed and agreed. What is the deal with mother Medicare anyway? When will they ever let the doctors make their own decisions? For new pump users visiting the Endo 4 times for the first year is reasonable, but when the patient shows the ability to use the pump correctly Medicare needs to get out of the way. Talking about a waste of money.

I just got my latest a1c (6.9) and it was much higher than the Clarity estimate of 5.9 for the past 30 days. So I’m wondering what accounts for such a big difference between estimated and actual values. My calibration fingersticks are usually within 10 points of the CGM readings (but not always). I also have been lax about checking BGs before meals and need to do more to avoid post-meal spikes.

BTW, my PCP was happy with the 6.9 but I’d like to get to 6.5 or better. I am meeting with my CDE next week to look my numbers, I:C ratio, diet, etc., to develop some strategies to meet my goal.

I also have a disconnect between my A1c as estimated by my Dexcom data. Remember that the Dex CGM is calibrated by your fingerstick meter. Have you checked your fingerstick meter against a lab draw? If your meter is close then that provides some better confidence that your CGM is relatively close to actual BGs. When I go for a blood draw, I carefully (with washed and dried hands) take two fingersticks and record them on my phone. I use the average of those two fingersticks to compare against the lab value that’s published a few days later.

My meter compares favorably to the lab and I still get an A1c that appears to be about 0.5% higher than my Clarity report estimation. I’ve puzzled over this and here’s the only possible explanation that makes sense to me. The A1c test is based on assumed longevity of red blood cells. Perhaps they took an average of a large population to build into the formula. Human physiology is not exact from person to person. What if your and my red blood cells lived appreciably longer than the average life-span baked into the A1c test? That would mean that our longer-lived red-blood cells are exposed to glucose for longer than the A1c test assumes. And that means that these red blood cells gather more glucose than if they died at the time the test is predicated on.

So, if you and I occupy the statistical position on the skirts of a bell curve for red blood cell lifespan, then the test is going to be less accurate for us. Consequently, people whose red blood cells die off more quickly than the test assumes will have A1c’s better than actual. I’ve read in the medical literature that certain conditions are recognized as skewing the A1c result.

In summary, I would strongly recommend that you find a meter that compares favorably with a lab draw for blood glucose and use that meter to calibrate your Dex. Also make sure that every calibration that you do is done with clean and dry hands. Remember that the purpose of the A1c test is to get an idea as to the all the tissues of the body’s relative exposure to blood glucose. If our red blood cells live longer and skew the A1c test results, it just means that A1c test overstates reality.

In any case, since I know that my meter compares well with lab glucose, I take comfort that the CGM numbers as displayed in the Clarity report are probably closer to actual than the A1c test.

Tomorrow, I will get my first A1C reading after being on a CGM for a full three-month period. Clarity would indicate it should be in the 5.5 area. I am very curious to see what the actual number is. If it’s closer to 6.0 (my last reading after being on the CGM for a month was 6.1, down from 6.4), then I might change the code on my Omnipod Freestyle glucometer from 16 to 18, which is what I used prior to getting on the CGM.

Just got my most recent A1c - 5.3. The last 30-days from Dexcom Clarity estimates 5.0. In my case, though, I thought I understood why there is a discrepancy - Clarity gets its numbers from the receiver, which I carry MOST of the time, but there are some occasions when I don’t - and some of those might have tipped the scale a blt – HOWEVER, the CGM data from my pump - which I carry all the time - gives me a BG average of 100, which translates to an A1c of 5.1 – a little closer, but still lower than the lab results.

I also think that the newer algorithms from Dexcom err on the low side. I find that the monitor frequently reports numbers that are lower than my meter and almost NEVER higher.

That’s a terrific suggestion, @Terry4, and I am going to put a reminder in my phone for my next draw (end of May). Thanks!

I did a little reading on this and it sure seems likely. I was dealing with a lot of stress in December and January and that affected my BGs a lot. My control has improved a lot since then. Another topic to discuss with my doctor, I reckon.

Well, I got my A1C, and it was 6.1, exactly what it was three months ago. Dexcom Clarity was giving me a pretty consistent reading in the 5.4-5.6 range, so I am pretty disappointed, although my endo was not unhappy at all. I was sure I’d at least be below 6.0. I am changing the strip code on my meter to 18. My strips are coded 16. Hopefully, that will give me a more meaningful number for calibration and a more accurate A1C estimate from Clarity.

@Jim2 – Have you ever checked your meter against the lab? I just had my A1c lab blood drawn yesterday. I’ve also experienced results like yours, about a 0.5% difference between Clarity estimate and actual lab A1c number.

Clarity was .5 lower than actual A1C from lab work.

I just received my lab results from a blood draw a few days ago. The A1c came in at 6.2%. The Dexcom Clarity report estimated A1c was 5.1% for the 14-day report and 5.2% for the 30-day report. The Dexcom Studio 90-day report recorded an average BG of 109 mg/dl which translates to an estimated A1c of 5.4%.

I did two fingersticks at the time of the blood draw with the meter that I use to calibrate my Dexcom CGM. They were 88 and 85 mg/dl or an average of 86.5 mg/dl. The lab blood glucose was 87 mg/dl. So the lab confirms that my meter tracks my actual blood glucose well. I’ve done this before over the years and have had similar results at a spectrum of BG levels.

So, Dexcom does not predict my A1c very well. As I’ve said before, there are a few explanations for this significant number disparity. One is that my red blood cells live longer than the assumption on which the A1c test bases its calculation. Another is that there is some confounding chemical presence in my blood. I don’t have anemia. Besides, that would explain an unexpectedly low A1c. I’ve asked my endocrinologist about this and the best I get is a shrug.

I consider my A1c as a false high. I believe my Dexcom CGM more accurately shows my exposure to blood glucose over time.

Based on the original paper where the correlation between A1c and average BG was established in a statistically significant manner, your results would not be super unusual at all. Here are the key graph and the related table:

In the table, the corresponding average BG is shown for each A1c. This is what doctors would readily quote or read from a chart. To dig a little deeper: the numbers in the parentheses show the range of average BG for 95% of the people with that particular A1c. For example, 95% of people who have A1c of 6, have their average BG between 100 and 152, which seems like a huge range. Things are not so bad. It is important to note that not all values in that range are equally likely. In the A1c=6 example, the standard deviation around the average BG of 126 is approximately 14, which means that almost 70% of people with A1c=6 have their average BG between 112 and 140, and more than 50% of them have the average BG between 116 and 136, a much narrower range. This is why many people will report that their A1c predicted by Dexcom matches the A1c lab test very well while some, like @Terry, may see larger differences. In any case, the relationship between A1c and average BG should be understood as a good approximation, not as a strict law,

Given your CGM data, and (importantly) the meter that seems to perform well against lab BG tests, I would consider your A1c test result irrelevant, as far as your BG control is concerned. However, it may (as you’ve noted) point to some other issues that would be worth investigating, unrelated to your BG control.

1 Like

Thanks, @Dragan1. I never had any formal training in statistics. It seems misleading to me that the “estimated average glucose” derived from an A1c number should not be expressed as a discrete number but instead as a range of numbers. For instance based on the table 2 you’ve shown above, an A1c of 5.0% means your average glucose ranges from 76-120 mg/dl, a significant spread. Conversely, an average blood glucose of 100 mg/dl could equal an A1c of 5.0% to 6.9%, a huge range to me.