Doc response

I’m not sure if I should calibrate. I’ve heard good arguments on both sides re: calibration. For sure, I calibrate if it’s, like, 100 points off.

Do You?

Early on I calibrated on day one and thereby destroyed the factory calibration. Now I just tolerate the lesser accuracy on day one. I am still waiting for a sensor that is off by a lot on day 2. Thus I haven’t calibrated in months.

1 Like

I haven’t had a doctor tell me A1c doesn’t matter, but would be pretty skeptical. There does seem to be more focus on time-in-range now although I don’t think that is based on long-term health outcomes as much as it is based on avoiding extreme lows.

The question of accuracy seems a little more nuanced. In my experience with meters and limited experience with my Dexcom G6, I don’t think I would characterize one as “more accurate” than the other (at least after the first day). I haven’t noticed consistent biases for either tool and the degree of uncertainty in any reading seems pretty consistent. The only other way I have to judge this is with my body which is still pretty accurate in telling me I am over 200 or under 70. What is (obviously) dramatically different is the number of data points. In the past, the average BG readings from my meter would underestimate my A1c by 0.2 to 0.4 because I took more readings right before meals and fewer in the 2 hours after meals. It’s not a meter accuracy issue, it’s how representative the samples are. The estimated A1c using CGM readings seems a lot more accurate, but I haven’t had an A1c since I started wearing the CGM.

You also raise an interesting question on driving privileges. In my state I have to submit a form signed by my doctor that indicates I’m “ok to drive” and recommends a duration. My old doctor would sign it for 4 years (max allowed), but now I have a new endo that I’m worried will make me deal with this more frequently. It’s a hassle with schedules, deadlines, etc.

1 Like

Ahhh, yes, @Bryan. I would agree with that. That’s an excellent way to phrase it.

Anybody know if Diabetes Technological Society has performed any verification of sensor accuracy, like they did for meters? It would be nice to have some general summary statistics for the population at large.

My endocrinologist also told mat I should aim to have 4% or less of my readings low. For me, that’s all but impossible even with an accurate sensor (Dexcom G5 was very accurate for me). I’m not sure I’ve been ever been able to hit that target. However, I don’t have a driver’s license to worry about, so I’m more curious than scared of how he’ll react when he sees 30% of my readings are “low”…

@Jen, what do you consider low?

The Libre will not let me set the low limit below 3.9 mmol/L. So it’s counting anything below that level as low. The problem is that it can read 3.6 mmol/L when my actual blood sugar is 5.0 mmol/L. This is why my A1c ended up high—I was correcting that type of “low” without actually checking on a meter. Now that I’ve started checking on a meter, well over half of what it claims are lows are true lows. (If my blood sugar were actually 3.9 mmol/L, I would probably treat at that level as it’s apt to drop further.)

I would only be concerned about below 3.0 mmol/L. Is there a way to calibrate the Libre to get rid of the low bias?

@Jen One of the ways to alleviate some of the questioning about the readings might be to do a bunch of finger sticks for a couple of days and write them down right next to the CGM reading so he can see the difference. That way he can visually see the comparison of how far off they are.

1 Like

That’s a good idea. I am testing fairly often, so he’ll be able to see the readings from my meter alongside the CGM graph. He didn’t seem to be the biggest fan of the Libre when I switched from the Dexcom to the Libre, so wouldn’t surprise me if he knows it’s less accurate for some people.

1 Like

@Jen Then you probably won’t have any issues, I know my endo was much more enthusiastic about the Dexcom’s accuracy. But it will help him see it when he can see the comparisons.

Just posting an image listing the conversions between mmol/L and mg/dL, so I can follow what you guys are saying…
3.0 = 54
3.9 = 70

image

1 Like

Anybody got fingerstick data for a Libre? I’m collecting for Dex G6, but we will see differences person to person, right? You got any data @Helmut?

@Jen, it will be a cold day in hell before I hit 4%. 8% is my absolute best. Note: thats just the 54 - 70 range that concerned him. If you add the “less than 54” numbers, it goes up to 12%. Thats the best I can do. I do approach 20% at times, I’m certain, If you have increased sensor error, that could bring it up around 30%, especially if your running a good a1c. My a1c currently sucks ■■■.

I have no data. I am past that point. I decided that G6 without calibration works for me. If I didn’t have a stockpile of BG strips, I would never finger stick again.

2 Likes

I think this shows error, for me, wherever blue dots dont hit the black line.
I think it shows that manual sticks definitely read higher than sensor because the dots fall above the line. It may also confirm greater accuracy at smaller numbers because none of the blue dots hit the black line for larger numbers.
So, perhaps it just confirms things we already knew. Worth a try.

image

Which is x-axis and which is y? Assuming x (horizontal axis) is the CGM and y (vertical axis) is the meter it shows a disturbingly high bias for the meter (or low bias for the CGM). As you note, they should center around the x=y line you have plotted. You could do a curve fit through the data. If it is a straight line with 0 intercept, the formula would look something like x=0.75y which is the correction factor you would have to apply to convert from one reading to the other. If you want to get really fancy, you could fit other types of equations. This would address the issue if high readings seemed to be more biased than low.

In addition to the bias, there seems to be a high degree of uncertainty in one or both of the measurements. If you assume the x-axis is accurate and look at one point, it looks like the y can vary by up to 50%. That’s extremely high! That would be like taking a measurement with your meter and getting 50 mg/dL and then immediately retesting and getting 150 mg/dL.

The biggest problem is that you have to start off with an assumption that one reading is accurate. Maybe neither are!

I haven’t done a detailed comparison of my meter and CGM, but I do at least one finger stick per day. Typically the results are within 5% (except for the first day I have the CGM on). I have retested many times with my meter and have rarely (maybe never?) seen a retest that was more than about 5% different from the first meter test.

It reminds me of the Confucius saying: “A person with one watch knows what time it is…a person with two is not sure!”

3 Likes

x is the sensor, y is manual stick. Your summary jives with my understanding.
I think the meter reads high and the sensor reads low.

The fact that there could be uncertainty in both devices is a real problem. Experience would tell me to blame the G6 because I simply never had this level of error with the G4. It was pretty spot on - with the exception that you had to take LOW readings with a grain of salt. G4 didn’t have so many false positives in the LOW area.

Thanks for the input, its helpful. I wish I could see some other people’s data. Theres some papers published. I’ll look at them today.

Thats what I have found - upwards of 50% difference in sensor from manual stick.
It might be presenting real problems with system reliability.

Look how it makes the categorization of readings in the tidepool output bounce around.

@mohe0001, did you calibrate the sensor for which you collected the data?

Confucius was a remarkable visionary since he made this statement 2000 years before watches were invented.

4 Likes