Dexcom G6 CGM trial

I have been using a Medtronic 670G insulin pump together with its Guardian 3 CGM since Feb. 2019. Many diabetics who use such systems have complained on Internet forums about this system and have praised others. When I asked my endocrinologist about the relative merits of the Medtronic CGM and the Dexcom G6 CGM, she offered me a sample Dexcom G6 CGM kit to try out. Later I was sent a second G6 sensor by Dexcom.

Over the past month or so, I have worn a Dexcom and a Guardian 3 CGM. I have about 25 days of tabulated data comparing BG from fingersticks* with readings from a Dexcom G6 and from a Guardian 3. I haven’t yet found a good way to display all the data concisely, so in this report I will write about the main things I saw, and end with a few subjective comments.

G6 Startup
I installed the Dexcom G6 app on an iPhone 8 (now running iOS 14.8) to read out the G6 sensors. At 5:15 pm on 8/31/21, I started the transmitter and the first Dexcom G6 sensor with its 4-digit factory code, initiating the 2-hour warm up. At about 7:50 pm the iPhone beeped and its home screen showed a high-glucose alert. The Dexcom app read 325 mg/dl while the Guardian 3 CGM showed 263. I measured my BG by a fingerstick – it was 261.

So the G6 was reading 64 mg/dl higher than the BG measurement. According to the webpage, a BG reading of 260 could show up on the G6 in the range 208 to 312. (This is +/- 20% – part of their so-called “Rule of 20”.) The G6 reading was 25% higher than the measured BG, while the Guardian G3 was in almost perfect agreement. I asked the pump to give a corrective bolus of insulin and both CGM readings came down over the next hour, but the G6 continued to be about 63 mg/dl above the Guardian 3 reading.

Calibration of the G6
I decided to calibrate the G6 even though Dexcom makes a big thing of saying it’s generally not necessary. (How would I have known whether a calibration was necessary if I hadn’t also been wearing the Guardian 3?)

At first I couldn’t find instructions on how to calibrate the G6, so about 9 pm I called Dexcom’s 24/7 Global Tech Support and waited 15 minutes for an agent to answer. While waiting, I found the instructions, and at 9:13 pm I did another fingerstick. My BG was then 229. The G6 read 293, about 28% higher than the BG, while the Guardian G3 read 221. I applied the BG calibration to both CGMs. Over the next 5 minutes the G6 reading dropped, ending up very close (233) to the BG measurement of 229.

The Dexcom 24/7 agent said I shouldn’t expect perfect agreement with BG and referred me to the “Rule of 20”, but said she would ship me a replacement sensor (perhaps because the G6 had been more than 20% too high).

Tracking the Dexcom G6 against BG
Overall I found that the Dexcom G6 sensors did well except for five things:

(1). As described above, the first G6 read too high by 25 or 28% until I calibrated it.

(2). At 9 pm on 9/02/2, the Guardian 3 read 57 and gave an alarm for low glucose. I was definitely feeling hypoglycemic. My BG measured 69, an uncomfortable level for me. However, the G6 was reading 82 and it gave no alarm of an existing or impending low – not good.

(3). On 9/09/21, 9:52 pm, I re-started the first G6 with its factory code (no BG calibration). Over the next three to seven hours it read too high (by more than 30%), similar to its initial behavior. However, all later readings fell within the “Rule of 20” even though I didn’t re-calibrate. This inconsistency didn’t seem too bothersome as the sensor was past its rated life.

(4). Owing to a delay in delivery of the second G6 sensor, I couldn’t install it until 9/17/21. I didn’t apply any BG calibrations as I wanted to track its performance using only its factory code. Over the first 30 hours the G6 deviated a great deal from measured BGs, reading as much as 40% too high and 19% too low. After this initial period all the G6 readings were within the “Rule of 20” except for the sixth day, when it read 30-34% high during all three of my BG measurements during one 24-hour period. I am not aware of anything I did to make the G6 readings go so high for one day and then return to acceptable values later.

(5). Both of these G6 sensors gave untrustworthy results on the first day and again after the eighth or ninth day. Re-starting the first sensor gave me some readings well beyond its 10-day rated life, but they seemed erratic and were interspersed with hours of no readings at all. I didn’t try re-starting the second sensor.

If the Guardian 3 hadn’t made me realize that something was wrong with certain Dexcom G6 readings, I would have been overtreating highs and missing lows. I find this worrisome. Were I to switch to the G6, I would be sure to do a fingerstick with each new sensor and perhaps again at the middle and near the end of its lifetime.

The sensor wire in the G6 is hair-thin and I was seriously concerned about it breaking off and remaining lodged in my body. Some correspondents at TuDiabetes have spoken of such an occurrence. The Guardian 3 sensors are much more substantial and seem to pose no risk of breakage.

In my data, the Dexcom G6 doesn’t show better accuracy compared with BG than the Guardian 3. Of course, the latter system requires frequent calibrations with measured BG, so the agreement of the Guardian 3 readings with BG may be due to the many fingersticks it requests (although I sometimes cheat and supply a BG guess rather than stick my finger!).

It must be admitted that the Medtronic 670G system has problems and annoyances. It demands frequent BG measurements, thus many fingersticks. It is very fussy and drops out of Auto mode at slight provocation. For example, if the glucose level hasn’t required any basal insulin for 2-1/2 hours straight, it sounds an alert, jumps into Manual mode, and demands a BG measurement to be returned to Auto. Or if the CGM detects a glucose level out of range (below 70 or above 280 mg/dl), it kicks the pump out of Auto mode. This has the result that the pump may not be adjusting the basal insulin just when such action is most needed. These too-strict criteria for staying in Auto mode cannot be modified by the user. I have read that the latest Medtronic pump and CGM system, the 780G, alleviates these problems, but it is not yet approved for sale in the U.S.

For me, the all-in-one nature of Medtronic systems is a real convenience. The Dexcom integration with a smartphone posed some difficulties for me. In my house the iPhone/Dexcom readout often gave a no-data alert when I left it in one room and stayed in another room for more than 20 minutes. When I leave my house I don’t like to carry a smartphone – I use instead a dumb flip phone that fits easily in my pocket; losing or breaking it would not be a major disaster. Dexcom’s G6 receiver appears smaller than a smartphone, so it can probably be kept in a pocket and thus always be within Bluetooth range of the sensor. It might be a good alternative for a person like me averse to carrying a smartphone everywhere.


That is very interesting reading and thank you for taking the effort to share! I’ve been using G6 for the past year and initially was worried that it may not be accurate so was still checking with finger pricks but did not calibrate the phone and found unless levels were changing, all results were within 5% over several months (and then my frequency of testing dropped off).
Lately I’ve been checking on random occasions. The biggest differences I noted ongoing between blood and G6 related to rising or dropping glucose levels; when the arrow was straight up, or straight down, the levels were often more than 20% off. Double up or double down: forget about it.

I find the dexcom is very accurate in normal range. It’s not so much when high.
I’m not at all surprised by your data.
Off topic, a 300+ reading is very high this will cause both sensors to crap out early. On top of all the other diabetes related issues.

The only time my sugar goes that high is if I pulled my infusion set completely out, still the cgm will start warning you at 160.

I glad the guardian works for you, it didn’t work at all for me.
Maybe individual chemistry has something to do with it.


So sorry you are having such huge differences in readings! This is the classic saying “What works for one person, might not work for the next.”
I find the same with pumps and even doctors. I like my doctor but there are some who don’t like him.
You just need to find what works for you and not worry about other’s opinions. I don’t worry whether my treatment plan doesn’t work for someone else. As long as it works for me, that’s all I care about.
So if your current pump/sensor system is working fine for you, keep using it. Just because everyone else says another system works great, does mean it will work for you.
There is no perfect CGM or perfect pump or perfect treatment plan. Everything has good and bad points. You just have to decide what might be a deal breaker.
Thanks for sharing your experiment! I did one once when doing a clinical trial. They use a YSI system that is a must when doing a trial. We compared all the different meters I have along with the CGM. I was very surprising. Love those fun experiments!


Perfect answer, @Sally7 , Off topic, what breed are the dogs in your avatar? I can’t tell, as someone who has lived and worked with Great Pyrs that’s what I see.

I wonder how good Medtronic sensors will have to become before some of us who used the original SofSensors would return to Medtronic. I was always quite happy with my Medtronic pumps but the SofSensors were painful, inaccurate, and not even accurate for trends. I used to get egg-size hematomas sometimes. I battled SofSensors on and off for 2-1/2 years starting in 2008 and was amazed when I moved to the Dexcom 7+ in 2011 that the Dexcom numbers and trends made sense. At that point I vowed to only purchase pumps that were compatible with Dexcom. I have erratic unreliable Dexcom G6 numbers on the first day but am very pleased with the sensor performance after that.


My understanding is the Dexcom is very accurate for 50-200, the higher you go the more it may differ from finger stick. Its very rare for me to be over 200 so I don’t have much experience with higher readings and how they differ from finger stick.

1 Like

If you ever get the chance to try another Dexcom sensor, you might also try “pre-soaking” it. It’s fairly common for the first 0-24 hours to be erratic. Those people who struggle with the first day blues, get around it by inserting the sensor [however many you require] hours before inserting the transmitter and starting the session. Gives the initial insertion trauma time to calm down before you start getting readings.

1 Like

Back in the days of Medtronic SofSensors, many of us presoaked sensors overnight. I am not sure whether it helped but it seemed a good idea.

I have presoaked Dexcom G6 sensors for 24 hours and have had no difference in my startup experience than just starting them right away. I think it is something with the startup algorithm that just reads low with my body chemistry. I have my best startup results when my meter BG is above 120. Interestingly when I used to restart G6 sensors, the restart always read high. I almost always calibrate my sensors 1-3 times on the first day and do great after that.

Having started my diabetes career with urine testing, I accept the weaknesses of current CGM systems and appreciate the life-changing technology.


all sensors no matter what they are sensing (engine sensors, fluid sensors, electrical sensors you name it), are calibrated to operate best in a range. Extreme readings high or low will not be nearly as accurate. It is virtually impossible to produce sensors that have accuracy at the upper and lower extremes. It is a value of accuracy and money.

To make them more accurate you have to do two things. 1. build in more redundancy (more sensors or points of sampling) and 2. make them more robust (last longer in extreme environments).

As they say in all instances, the issue is not much about if they can or not. The matter is it is just how much we are willing to pay for the improved sensor.

1 Like

That’s not entirely true, although I agree that they target normal range.
I think it is a limitation of the hexonase reagent. It can’t deliver enough current when it surpasses 300 mg/ dL. It flattens out and more glucose will not give a linear predictable rise.

But when any assay is calibrated, the normal range is set. And there are several points within that range.

So on cgm usually it’s 50 80 110 140 170.
Taht makes for. 5 point linear curve. And it is accurate from 50 to 170. As you get outside the range the accuracy suffers.

In the case of cgm it suffers a lot in the upper range because the technology just can’t be linear at high concentrations.

There is only so much reagent on that wire, and it gets used up.
That’s why sensors usually crap out early if you run high a lot


My two little puppies are yellow labs! Big one is 5 and the smaller one is 3. They are sisters. Got them from a breeder in Riverside county in Southern California. They are the English labs, so a little shorter, stockier and can be pretty lazy. Unlike our first lab that was an American type. Much taller, thinner and never ever stopped! Love our little labs! The kids drove us crazy for a dog and labs won out!

1 Like

That’s interesting, we saw labs like that in Ireland, several were lounging in pub doors, so much so we started calling them pub dogs. They were always friendly pups.

I agree with you. The optimum curve for a sensor is 50-170. At the extremes it is more wonky than in the middle. If the sensor were designed additional normal range the sensor would cost more because it would need or something. More regents? More sensors? Just more of something. That would trade cost, efficiency, then number of changes, something.

It is the truth for all sensors no matter what it is sensing.

Timothy thanks for educating me on the inner working of CGM’s. I needed the education


1 Like

It may be true that sensors work best between 50 - 170, but I am seldom out of that range and I have had plenty of wonky sensors.

1 Like

I’ve only had the G6 for a year. I’m sorry it didn’t work out for you. I found it confusing in the beginning, but now I can’t imagine life without it.

I was finger sticking over 8 times a day as I was originally misdiagnosed with (T2 vs LADA) and new to insulin dosing.

What I learned/like about the G6 is:

• positioning the sensor is critical for accurate readings. I now alternate between two spots and that’s it.

When I originally tried to alternate around the belly area as recommended, different spots and/or the sensors would produce more fails or wildly erratic readings.

• Smartphone app integration and sharing readings. One device. On all the time. So easy. Plus, being new to lows, having someone else receive my data and act as backup if alerted was critical.

• customization on settings. So many beeps. :slight_smile:

• real-time understanding of exercise, foods and their impact. Or, in the early days when I was still learning how to inject my insulin, seeing if I did it correctly. Being out hiking and getting a low reading. I’ve gone low so many times without understanding what it meant. Life saving.


Like @Mika, the pluses far outweigh the minuses for me. No more lows!! Early notifications of trends! Control of the pump (that I got only after eight months with Dexcom)!
It isn’t like not having diabetes, but it has made a huge difference for me in terms of A1C (and time in range).
I put the sensors on my upper, outer side of triceps, alternating each 10 days.

Well yes. There are a lot of variables that go into this thing.
I have about a 10% failure rate with no real explanation.

I seldom get 10 full days out of a sensor. I get woke up by, I forget, signal loss alerts, usually around day 7. It tends to get worse as I get to day 9, losing signal for up to 3 hours as I am up and about.

I don’t have extreme highs and lows, been 99% time in range of between 70 to 145.

I wonder if individual body chemistry or perhaps the immune system can affect the reagent and/or wire?

I really would like to have a conversation with some person on the design team instead of Dexcom telephone reps with limited knowledge.

Everybody’s body chemistry is so different with sensors. The Medtronic sensors never worked for me. They were a nightmare for me to attach; for my brain, counter-intuitive to even pull off the paper covering the sticky pad. When I did get a sensor attached, they were never accurate. The Freestyle Libre is never accurate for me either, so I wasn’t all that surprised. I mostly love the Dexcom G6. I always have wonky readings the first 18 - 24 hours if the sensor batch starts with 59. I know I’m going to have issues as soon as I see that number. But, eventually, the readings settle down and are the most accurate for me, compared to other systems. I’ve never had one stop working early, although I have had ones that wouldn’t work at all. Dexcom will send me a replacement quite easily when I call them about it. Medtronic always wanted to blame me and wouldn’t send replacements. My blood sugars only go higher than 12 if I’ve bolused incorrectly, or if my infusion set has been tugged on, or if a newly inserted one simply isn’t working. If that happens, I use finger sticks to verify my readings. The BG sensors don’t like very highs, or very lows. Anyway, long story short, use what works for you. I think as time goes by, they will work on improving accuracy with all the sensor systems.