I am new to pumps, recently starting on an Omnipod 5 with a Dexcom sensor. I thought I was doing okay, showing 6.7 GMI with 85% TIR, but my A1c just came back as 7.1. I am thinking the solution is a daily calibration to make sure the sensor is accurate.
Are there better solutions, and how do others keep their sensor accurate?
I don’t have an answer for you but have a similar problem. My A1c is usually 5.7-6.0 and my GMI is usually 6.7-6.8. it’s a different calculation altogether so I just try to keep my average BG down as much as I can and don’t worry too much about the longer-term stats.
Edit - I don’t think calibration is the answer if you are checking with a meter and the number matches with your CGM. Obviously if it doesn’t match, you should calibrate.
GMI is based upon sort of an average of blood glucose measured every 5 minutes. HbA1c is the percent of hemoglobin that has been glycosated becoming hemoglobin A1c. While GMI is an average of BG, HbA1c is not. Worse the GMI is an estimate of HbA1c.
It’s kind of apples and oranges. The 2 metrics do not measure the same thing. The spread between your GMI and HbA1c are not way out of bounds. For some people they will be the same or a point different, but not all people.
This image shows the possible variables of an HbA1c of 7.0% correlated to time in range. I am more concerned about being in range than HbA1c numbers.
(upload://7cT5AyKFMbJdZTxuOzrVL0mM93H.jpeg)
Here’s a screenshot of my last 14 days from Dexcom Clarity app.
The standard deviation is how much my blood glucose was above or below the average. BTW my last HbAic an month ago was 6.4% the same as this GMI.
I don’t think your issue is inaccurate sensors. A lower A1c comes from a lower average BG, especially in the few weeks before the A1c measurement. To get a lower average BG requires more insulin. More than the Omnipod 5 will give by itself, because the algorithm won’t risk driving the BG low, whereas we can take larger manual doses of insulin to curtail a BG rise that may result from an increase in insulin resistance or to compensate for impaired insulin absorption or leaking at the cannula. We can take a larger insulin dose because we can take fast carbs if/when it turns out that we took too much.
A way to lower A1c is to limit BG rises by more aggressive dosing, while being prepared to counteract the BG drop with fast carbs when too much insulin has been taken. Don’t worry about the difference between GMI and A1c, just work to reduce the magnitude of BG excursions and the BG average.
@James.J.Igoe I have to agree with @allison in that daily calibration is probably not the answer. I think for most folks one or two calibrations per sensor (that’s 2x two calibrations with about 5 min between them each) is about the most I find acceptable for a G7 before replacement. My recommendation is to rely on TIR (TITR if so inclined), GMI, and COV with SD. If those are reasonable, A1c will most likely be reasonable as well, though it won’t be the same as GMI. @CarlosLuis’s charts give the right reasons (wide swings in both low and highs will give a false impression of a stable acceptable number) plus A1c is looking backwards about 3 months stating where you’ve been while the others can be adjusted for currency.
The most likely indication is that your HbA1c is going down now you are on the O5. HbA1c gives a weighted average over the last three months so if you have been on the O5 less than three months you are seeing history in your HbA1c which is not reflected in Insulet’s “GMI” estimate.
I use Dash, AAPS and Dexcom7 but my HbA1c is always significantly lower than the estimates from the various estimators (xDrip+ and AAPS in my case) The estimates also differ themselves, ATM:
xDrip+ 7.4%
AAPS 7.1%
Partly that reflects the Dexcom; for me it seems to read high a lot of the time but not always. It certainly is not accurate (I’ll explain what that means). I just fingerstuck, the Dexcom says I’m at 211mg/dL the fingerstick says 226mg/dL.
The permitted range is +/-20mg/dL below 100mg/dL and +/-20% of the actual value above 19 times out of 20. Dexcom is well within range for me when I’m above 100mg/dL by the fingerstick, but for fingersticks the permitted accuracy is only slightly better (15 instead of 20).
The inverse of GMI for HbA1c is called eAG - “estimated Average Glucose”. Here’s the skinny:
I work better with algebraic formulae than with tables, this is the formula:
28.7 X A1C – 46.7 = eAG
So each 1% rise in HbA1c corresponds to a 30mg/dL (28.7 from the formula) rise in average BG! Check out the table if you don’t do algebra.
That means that the errors in a Dexcom, +/-20mg/dL, amount to a 1.4% difference in HbA1c. Of course the errors average out somewhat, but in a short term average (like you have) the GMI is likely to be massively off.
To make this worse, much worse, the programs I have do not do a true HbA1c calculation; they seem to just average and I think xDrip+ is using a very small timescale. HbA1c does not change over one day or one week; it changes on a month-by-month basis because it represents the average over red blood cell life, which is around 120 days (though not for all people.)
HbA1c is my holey grail - it factors out the errors in the CGM completely and relies only on the lifetime of my own red blood cells. My doc gets one every three months and if it was as high as xDrip+ or AAPS suggest he would get whiney. As it is (around 6.5%) he doesn’t want me to change a thing.
GMI is an average (unspecified) of inaccurately measured blood glucose. HbA1c is a time weighted average (so exactly specified) of red blood cell glycosylation; the rate of increase of later is directly related to the level of glucose in the blood around the cells.
So GMI reflects measurements by some arbitrary BG monitor averaged in some unspecified way.
If done correctly GMI uses a time weight over the last 120 days. At time t(0-120 days) in the past BG was b and the red blood cells alive then got glycosylated at an instantaneous rate proportional to b. For a red blood cell about to die that’s a simple integral over 0..120 days. Then the current glycosylation is the result of integrating the red blood cells over their youth, so to speak; assume red blood cells are produced at a constant rate and die after exactly 120 days (a simplification).
Ok, it’s not a simple time weight, but it is simple math.
Insulet can, and should, do that math; Insulet should get the math right and the average should be calculated correctly. Therefore the O5 GMI should reflect only the inaccuracies in the CGM for people who’s red blood cell life is whatever Insulet’s assumption is so long as the data has been accumulated over the maximum red blood cell life.
GMI really is meant to be an estimate of HbA1c. Performing an arithmetic mean on the last day’s readings is not such an estimate!
Based on your low GMI and high TIR, I think you ARE doing OK. Damage to hemoglobin in RBCs was easily measured in the old days, which made A1cs a convenient reporter of glucose levels, but hemoglobin damage from glucose has no special health significance otherwise. High glucose levels also damage nerve and kidney cells, leading to common diabetes complications, but that damage isn’t as easy to measure in a lab. My own RBCs are shorter-lived than usual (anemia), which gives them less time to accumulate glucose damage and gives me much “better” A1c values than I have any right to. My endo pays more attention to my actual blood glucose levels to judge how I’m doing while healthcare providers like opthalmologists and surgeons who went to med school before CGMs were invented are used to relying on A1cs. As long as you’re avoiding dangerous lows and highs, I think you’re doing better than OK.
That’s also a factor in HbA1c accuracy.
I thought my A1cs were affected by my routine blood donation, but I recently had a couple of months I couldn’t donate and my A1c was the same. That was a nice validation that my donations WEREN’T having as much impact as I’d thought!
If you were having higher glucose levels shortly before your A1C test, your A1C level could show a little higher result than expected.
Thanks everyone: @petesteve @allison @CarlosLuis @jbowler @mjf @TomH @bkh I do appreciate the input.
My feeling is that recent and extended high blood sugars, likely around how I manage my hour-long Zone 2 workouts, i.e., send it high (200+) and refuel with 25 grams at the 20 minute mark, so that by the end my blood sugar is within range, usually close to the lower bound.
I hope this isn’t too offensive, but I use the AI tools to get an overview. The one below is from Gemini in response to “reason 90-day gmi differs from hba1c”.
Gemini AI:
The Glucose Management Indicator (GMI) and a laboratory HbA1c differ because they measure completely different biological processes and use different calculation methods, even when evaluated over a 90-day window. [1, 2]
Core Reasons for the Difference
That’s actually a good summary by any reasonable account. So far my experience is that the Amphitheater’s AI summaries are typically better than most human summaries, my own included. It just comes down to reading more ■■■■ and knowing exactly what every human is reading.
@James.J.Igoe While I’m not an AI fan without a good strong analytical review, the summary provided sounds about right. Both GMI and A1c have their place, pros and cons. It’s your choice what to use and believe.
I’ve decided to take 2 actions to bring my GMI and A1C in line, or at least make sure neither number is high enough to trigger concern:
And you may need insulin afterward to avoid those post exercise highs (at least this happens to @Liam-M.)
A response from Gemini AI regarding my workouts and activity, but essentially, it suggests better use of Activity Mode on the Omnipod to avoid the need to start high:
Fine-Tuning Your Routine: Ideas to Consider
Because you already have a strong foundation, here are a few advanced nuances that might help you further optimize your blood sugar stability:
This sounds like the advice the Tandem trainer gave me when I got my pump. She suggested going to exercise mode 45 minutes before beginning the activity.
That’s probably good advice for many. It was not for me. In a 2 hour+ ride on exercise mode my blood glucose levels will just rise unacceptable high.
Before a ride I would have a relatively high BG from breakfast. With basal set to regular mode my BG would steadily but slowly drop.
Sometimes around the halfway point I would stop and treat an impending low. If not in hypo it would be 22g carb from peanut butter crackers. This would give a slow not extreme rise.
If hypo then glucose tables followed by the crackers to give sustainability.
My riding even with the pedal assist bike is in heart rate zone 2 to 3. When I was younger on a non ebike I would routinely hit zone 5, doing sprints and little impromptu competition with my fellow riders. At 76 those days have past.
Anyway try using the activity mode, experiment a bit with the timing to find what works for you. It might be 90 minutes or even 45.
85% time in range is great! A1c and GMI don’t always match perfectly, so check with your diabetes team before starting daily calibrations.
