Comparison of Continuos Glucose Monitoring systems (CGMs)

Which CGM is the best? At DiabetesLab, we analyzed your opinions, and here is a summary.

Accuracy of different systems

CGM accuracy is measured as MARD (mean absolute relative difference between CGM readings and blood glucose readings). This is an error metric — lower MARD is better.

  • Dexcom G5 (as well as Dexcom 505 AP in the USA): MARD 9% officially [1].
  • Dexcom G4 original algorithm: MARD 13.00% officially [2], MARD 12.60% in the test by IDS.
  • Medtronic Enlite: MARD 13.60% officially [3], MARD 18.66% in a test by IDS.
  • Abbott Freestyle Navigator 2: MARD 11% officially.
  • Abbott Freestyle Libre: MARD 11,4% [4].

1. J Diabetes Sci Technol. 2015 Mar;9(2):209–14. doi: 10.1177/1932296814559746. Epub 2014 Nov 3.
2. Diabetes Technol Ther. 2013 Oct;15(10):881–8. doi: 10.1089/dia.2013.0077. Epub 2013 Jun 18.
3. Diabetes Technol Ther. 2014 May;16(5):277–83. doi: 10.1089/dia.2013.0222. Epub 2014 Apr 7.
4. Diabetes Technol Ther. 2015 Nov;17(11):787–94. doi: 10.1089/dia.2014.0378. Epub 2015 Jul 14.

Dexcom G4 and G5

Dexcom G4 Transmitter
Dexcom G4 transmitter

With the latest algorithm update, Dexcom became the most accurate CGM system on the market. FDA-approved sensor life is 7 days, but most people would restart sensor when it expires it tends to track accurately on the 2nd and 3rd week.The most recent version of the system, Dexcom G5, is already shipping in the Unites States and the United Kingdom. There is a new app Dexcom Clarity which spots glucose insights. G5 has features such as displaying CGM data directly on a smartphone, without having to keep a Dexcom receiver around. Data can be transmitted to iOS devices and Apple watches (USA) or a variety of other devices worldwide thanks to Nightscout’s do-it-yourself projects. There are still things to tweak in G5; here is a review.

Current versions available on the market:

  • Dexcom G4 Platinum, a classic stand-alone system, which is the most recent one available in Europe and does not include 505 AP algorithm update. It doesn’t communicate with smartphones.
  • Dexcom G4 transmitting data to Animas’ Vibe insulin pumps. You don’t need a separate receiver — CGM data is shown on the pump screen. The algorithm is hardcoded into the pump and doesn’t include the latest update. No smartphone communication.
  • 505 “Artificial Pancreas” software update for Dexcom G4 released in 2014 in the United States — improved accuracy.
  • New Dexcom G4 with Share receivers available in the US. These receivers already work with the 505 algorithm and can transmit data to a smartphone.
  • Tandem t:slim insulin pump integrated with Dexcom G4 (US only).
  • Dexcom G5 with all the latest updates — already on the market in the US and is about to appear in Europe.

Medtronic Enlite

Medtronic MiniMed 530G insulin pump with an Enlite CGM sensor and a Minimed Connect device to upload data to a smartphone. Image credit: Medtronic.

Medtronic sells CGM sensors which connect to their insulin pumps. Stand-alone devices are also available. Medtronic sensors are statistically less accurate than Dexcom (MARD 13.6%), which can be frustrating for some users. Approved sensor life is 6 days; in practice people would restart it, to make 10–12 days out of sensor life.

Users’ opinions:

  • not so accurate;
  • uncomfortable to wear;
  • there are happy users, though!

Facebook screenshots with examples of users’ feedback:

Users' feedback on Enlite CGM sensors. Screenshots from Facebook.

This CGM is easier to try for Medtronic pump users: the initial costs will be lower, because the data is shown on the screen of the pump you already have. With a recently launched Bluetooth device MiniMed Connect ($199), pump and CGM data can be transmitted to a smartphone app, for now iOS only, and US only.

Abbott Freestyle Navigator II

Images credit: Sugarless

Abbott Freestyle Navigator II is available only in specific countries, such as UK, Netherlands, Israel. It is not planned to be available in the United States. Navigator was the most accurate CGM in the world before the launch of Dexcom’s 505 algorithm.

It provides a new glucose reading every minute, as opposed to 5 minutes for Dexcom. Approved sensor life is 5 days, but it may last longer. Navigator requires only 5 calibrations in 5 days, at increasing intervals, using a built-in glucometer. Occasionally the system may ask for additional calibration in case of unstable glucose levels.

There are predictive alarms based on glucose trends.

“My №1 problem with the sensor is the adhesive performance, or lack thereof. [solved by using a wipe and a Flexifix tape] “I also had a problem with a couple of faulty sensor inserters that didn’t release the sensor into the skin.”
“I have found the Nav2 CGM readings to be quite accurate, typically within 10% of the FreeStyle glucometer results. I have done some random tests to compare, and can say non-scientifically that I am very happy with the system’s accuracy.”
“Here in Israel I have full, literally 100% reimbursement so I haven’t had to extend the sensor life to save cash. As to cost, I asked last year and it was something like $2,000 for the device and $600 monthly for the consumables.”
Freestylin’ in Israel with the Navigator II CGM by Don Weintraub @ DiabetesMine

Abbott Freestyle Libre

Image credit: Abbott

Libre is in clinical trials in the United States and is available in certain countries in Europe. There are problems with its availability because of low manufacturing capacity.

It’s “flash glucose monitoring”, not real-time monitoring: you need to manually slide the device past the sensor to receive the latest data. No alarms — because information is not continuously transmitted from the sensor to the device. So we cannot call it a real CGM. Abbott markets the device as something in between CGM and fingersticks.

Libre doesn’t require a separate transmitter — everything is integrated into the sensor. Sensor is placed on forearm; device communicates with the sensor via NFC. Sensor stores up to 8 hours of information. You need to slide the device past the sensor at least once in 8 hours to save a complete picture of glucose. To save storage, Freestyle Libre keeps only one glucose reading for each 15 minutes, even though it takes measurements each minute.

Libre doesn’t need calibration — because it’s “factory-calibrated”. You can’t do anything with bad sensors, since you cannot calibrate them. Libre uses predictive algorithms to minimize delay with interstitial fluid. Then rewrites history if it was wrong.

“in many cases, the Libre predicts what your BG should be by projecting itself in the future. Sometimes, it gets it wrong, especially when for example it sees a constant increase and you start to exercise.” (Pierre)

Likely to use an algorithm to compensate for the temperature.

“Smaller, less painful sensor than traditional CGM. Very simple insertion process, does not require training. FreeStyle Libre is approved for dosing insulin except in three cases: when hypoglycemic, when glucose is changing rapidly, or when symptoms don’t match the system’s readings.” (DiaTribe review)

There is an option to use Libre as a traditional glucose meter — with Freestyle Optium test strips for blood glucose and blood ketones.

Libre also has a hidden bolus calculator, which is supposed to be set up by a healthcare professional, and it can be activated by entering a special code.

CGM graph on the screen is not so detailed, you cannot zoom it.

“It seems to me that the Libre has been designed around the needs of multiple injection people rather than pumpers who are more likely to be using “sugar surfing” techniques” (Facebook)

Libre is not approved for driving in the UK:

“Police & DVLA won’t accept readings from the Libre, you can be asked to show your BG meter if stopped.” (Facebook)

Libre readers are targeted at each specific country: units (mg/dL or mmol/L) are hardcoded in the reader. The same with the language. Currently you can buy an English version only with mmol/L (UK market).

A Libre sensor lasts for 14 days (hardcoded). The sensor cannot be restarted as a new one. The first 1–2 days are normally less accurate. It helps to insert the sensor 1 day before, so that the sensor starts tracking more accurately from the beginning.

Abbott put a lot of work on visualization techniques. Example of a “typical day” report:

Glucose insights of a person without diabetes testing Libre
Glucose insights of a DiabetesLab member who doesn’t have diabetes and was testing Libre (that’s why so many untreated lows!)

Experience with different systems

Simultaneous run of Libre and Dexcom G4:

The sensors weren’t perfect: the Libre missed a single clinical, meter confirmed, low on the third day. The Dexcom had several misses that we found extremely annoying. For example, the Dexcom missed a bad low on day 12. We wouldn’t run a totally automated artificial pancreas that aims for our level of control on either sensor alone. 
 — “14 days with the Dexcom G4 (non AP) and the Freestyle Libre” by Pierre Vandevenne

Difference between accuracy of Dexcom, xDrip and Libre is largely explained by the difference in the algorithms (not only the sensor technology).

“To put it bluntly, once you have used well behaved Libre sensors, it is extremely hard to remain satisfied with the performance of the G4 non-AP system. Our initial tests seem to show that [xDrip] performance is closer to a good Libre sensor (and to the 505 AP Dexcom algorithm should be) than the standard G4 non AP system. [..] I suspect it will be somewhat constrained in terms of speed by the sparse 5 min Dexcom sampling, but we’ll see.“
 — Abbott Libre Update by Pierre Vandevenne

Time lag of Dexcom (in comparison with blood glucose readings) is larger than time lag of Libre. Still both sensors are not too reliable in case of rapidly changing levels:

The only times when [Libre] has been wildly out has been when I have had a faulty sensor (which Abbott replaced free of charge) or when my blood sugar has been rapidly changing and this has been visible from the trending arrows. Sometimes there are gaps in the Freestyle Libre data, this appears to be when my blood sugar has changed very rapidly.
The only time I found the Dexcom to be inaccurate was when I became frustrated and tried to overly calibrate it. However once I left it to work itself out and have stuck to 12 hour calibrations the accuracy is great.
“Dexcom vs. Libre”

Our winner

From the point of view of accuracy, device connectivity and ecosystem, our winner is Dexcom. The decision is always yours — you may want to consider factors such as insurance coverage, out-of-pocket costs, or availability of specific devices in your country. A promising runner-up is Freestyle Libre which is more discrete to wear and doesn’t require calibrations, but doesn’t have all benefits of Dexcom yet.

Products coming soon

  • Senseonics Eversense, an implanted CGM that lasts 90 days. The system consists of an implantable sensor, an external transmitter, and a smartphone app. Exceptional results in a clinical trial: 8.8% MARD, better than any CGM on the market. Eversense is already available in Sweden and Germany, and scheduled to launch in other countries of Europe and Middle East.
  • Medtrum is a tubeless artificial pancreas initiative from China. They have a disposable CGM system, EasySense. Reported accuracy is at the level of Dexcom G5 (MARD around 9%). The company announced a distributor in the UK.
  • SugarBeat CGM system, with a daily-disposable adhesive skin-patch, has been approved in Europe. Its launch is planned for 2017.
  • Dexcom G6 may be available in 2017.
  • Medtronic developed a new more accurate sensor, Guardian 3, with reported MARD of 10.55%. Recently approved Medtronic’s MiniMed 670G pump (to be launched in Spring 2017) will use this sensor.

Last updated: 22 January 2017



I have used Medtronic CGM (Sof-Sensor and Enlite) intermittently (due to cost) between 2010–2015. Overall the results are satisfactory Some episodes of sensor failing (Sof-Sensor). Thank you for the CGM reviews, very helpful. December 25, 2015 at 14:16


lower mard high in accuracy why dexcom is the winner although it higher in mard. May 10, 2016 at 05:47


Dexcom G5’s MARD is 9%, the lowest one. October 14, 2016 at 10:30.


I have been a Dexcom costumer for years, and have raved about their products to anyone who would lend me their ear. That is why it is discerning that I am writing this today.
I have changed insurance companies a lot being an government contractor. Some of which I have had my durable medical supplies 100% covered. Over the past 9–12 months Dexcom customer service has twice told me that I did not owe anything when placing my order, even after I specifically asked about my 20% copay. 6 months after this happened the first time I got a bill in the mail for $384, yes my 20% for the 2 previous orders I placed. Should we pay a company when their representatives clearly tell us we don’t owe them anything? I am having trouble swallowing that pill, though I have never had this issue when I needed pills…
Planning to change permanent jobs and eventual insurance carriers, I called Dexcom on May 20th 2016 to verify my prescription and all other yearly paperwork is still valid so I could place my last order on this insurance at the end of May. When I called to place my order today, May 31st 2016, I was told my paperwork had expired on May 13th. with the time to get the paperwork updated and to also now do an insurance evaluation I would not get my order under this ending Insurance plan.
At this point the slightly superior product is out weighed by the now extremely poor customer service. I will no longer recommend Dexcom to anyone and thus will explain my experience to everyone.
May 31, 2016 at 20:43


My son used the Freestyle Libre, for a year, we loved it but eventually needed the alarms that full CGM offered. We’ve been using Dexcom G5 for 3 months now but there are still aspects of the Libre that I miss….most notably the simple insertion process but I love the remote monitoring and alarms on Dex..that is game changing.
June 1, 2016 at 08:40


My 12 year old is using Libre but I decide to try dexcom 4 since I really can uae the alrm function at night.
Was it difficult for your child to switch?
Best regards from Sweden
March 2, 2017 at 20:15 Reply


How do you connect a dexcom g4 with a medtronic 530g


There is no way. Medtonic’s pumps are not integrated with Dexcom. October 14, 2016 at 10:33


I would like to know which continuous glucose monitor ring system is the best and whether the European version is better than the American version.
June 14, 2016 at 02:16


If you can get Dexcom G5 in Europe, then it’s probably the best (and the same one as in the USA).
October 14, 2016 at 10:35


I hv used freestyle libre pro by Abbott India in 15 of my patients. It is very useful but it gives much lower readings as compared to glucometer readings by same company.
August 29, 2016 at 17:18


I wish I read this before getting the Medtronic 630g. With the Medtronic auto suspend feature, accuracy is important. And the Medtronic Enlite simply is not accurate enough to use with an insulin pump. When the Medtronic CGM says you are 80 and you are really 50 that means the auto suspended won’t be kick ing in and the Medtronic CGM will tell the pump to kill you with an overdose of insulin. Any and every time one computer is giving another computer instructions accuracy is important. My Medtronic has done this to me several times at work and in my sleep. Don’t let anyone ever tell you accuracy is not the most important criteria. Accuracy is a matter of life and death.
January 22, 2017 at 05:13


[…] […]
February 1, 2017 at 00:31


Absolutely go with the Dexcom. Enlighten is a complete disaster: up to 20% error margin in glucose readings, painful injection/insertion process, often misapplication or insert error which requires a phone call to receive a replacement sensor. Complete waste of time. The Dexcom *just* *works*. It is simple and painless and the error window is blow 9%. Nothing else comes close.
February 19, 2017 at 15:15


Out of testing 10 daily
What is the % of It being as Accurate as your Test meter?
And don’t give me CGM can be upto 20% error
I’ve had others who tried them, find they still ened up testing teh Same or MORe
after getting a Alarm that their BG’s were either too High Or Going too Low and testing manually? They werent!
So were are the Stats of say 100 CGM users that used to test 10x day ?
and How Accurate is it compared to a Meter?
and how about the OTC Prices and annual Costs? Without Insurace..Seeing as Most Insulin Diabetics are UnInsurable..
Thank you
June 8, 2017 at 11:02


The point of CGM is not being as accurate as test meter; the point of CGM is showing the glucose trend. CGMs and blood glucose meters measure different things: glucose in interstitial fluid vs. blood glucose. Those are correlated but it’s not the same thing. CGM sensor readings are 7–20 minutes behind blood glucose values, which is particularly important to keep in mind when glucose is changing rapidly.
Newer CGMs are more accurate than previous ones but don’t always replace fingersticks. For example, Freestyle Libre and Dexcom G5 are approved for insulin dosing but with a number of exceptions where fingersticks are required (e.g. when glucose is changing rapidly).
Dexcom G5 needs to be calibrated properly (and this is a potential source of errors, because there are rules for calibration which not everyone is aware of). Libre works without calibrations but is not available worldwide (e.g. it’s still waiting for the FDA approval in the United States).
For the statistics, you can look at the table in the beginning of this blog post and check out more clinical studies in MedLine / Google Scholar.
As of annual costs, this depends on the country.
August 3, 2017 at 06:33


Has anyone found out if innovations like the patches or contact lenses have been tested enough to be put on the market?
Both posts have a link to the original article for more info, for those interested.
August 18, 2017 at 15:06


The most tested one is probably SugarBeat, a patch for glucose monitoring which is already approved in Europe. It is expected to launch in the UK by the end of 2017.
The reported accuracy of SugarBeat is lower than ‘invasive’ glucose monitoring but is still decent enough, at least for type 2 diabetes and pre-diabetes:
> the results indicated a MARD of 14.05%
Their press release:
There are many other projects. I might write a separate blog post about it. For now, you can take a look at my Quora answer with several examples of non-invasive glucose monitoring, e.g. there is a glucose-monitoring smartwatch company which is actively raising money to launch in 2018:
August 19, 2017 at 10:14


I’ve used Libre for 15 months. The +/-9% accuracy is not an issue in my opinion. The lag time is also not a major disadvantage. High and low blood sugars can be pre-emptively avoided. The trends are clear even if they lag. It also has to be understood that hypoglycaemia impacts us when the brain cells are not receiving sufficient glucose via the interstitial fluid.
Another point is that since the introduction of stable long acting insulins, insulin pumps are arguably redundant. In my opinion the only benefit they offered was drip feeding fast acting insulin to simulate stable performing long acting insulin. Bolus shots are easy to administer. So why use a pump?
January 12, 2018 at 09:41


Dexcom G5 user; unit works well except when exercising hard(heart rate above 80% max) readings come in extremely low until I finish exercising. Spoke with dexcom rep’s no explanation??
Example: dexcom G 5 reading 50 blood glucose monitor reading 150. Great unit not for competitive exercise.
January 17, 2018 at 00:47


Weird. Dexcom readings are around 15–20 minutes late compared to fingerstick readings — does this huge difference still remain even if we take the delay into account?
January 17, 2018 at 10:19
Like what you read? Give Olga Pustovalova a round of applause.

From a quick cheer to a standing ovation, clap to show how much you enjoyed this story.