People often ask us: “Are continuous glucose monitors (CGMs) accurate?” Our simple answer will typically be "yes", particularly if we are short on time. But, if you want to understand a little more about what accuracy means, and how it is evaluated, in the context of CGMs, you should read the rest of this blog...
"Accuracy" is a term used to describe if a measurement tool, like the pulse rate monitor on your high-end expensive watch or your chest band heart rate monitor, truly represents the actual value of something (like your heart rate).
For example, let’s say that you are cycling at your steady state Functional Threshold Power (FTP) in an exercise lab, and your actual heart rate, measured using an electrocardiogram (ECG), is 129 beats per minute (bpm), but your expensive wristwatch with a pulse rate monitor indicates 170 bpm, then you clearly have an accuracy problem with that expensive watch.
Challenges in CGM accuracy
Do accuracy problems exist with CGMs?....Yep, they do….at times. Some CGM units might be more accurate than others, and for some people, for unclear reasons, and in some situations, CGM accuracy appears to deteriorate. But, we need to explain a few key terms before we delve deeper into why this may happen.
Some definitions worth knowing
The term "accuracy" is often overused. From time to time, we should be using words like "precision" and "reliability" in addition to “accuracy” when we talk about our exercise wearables and other gear.
Accuracy (validity): The degree to which a measurement represents the true value of something. Simply put: How close a measurement is to the true value. Accuracy is assessed by taking several measurements with a new tool or device, and this cluster of data is then compared to a so called “gold standard” (a tool that is thought to be the most accurate for that same measurement). Since a CGM measures interstitial glucose every few minutes (as we discussed in a previous blog), you might wonder how accuracy can be assessed. For CGMs, assessing accuracy requires numerous blood samples, usually obtained with a venous catheter, and lots of data analytics.
Precision (reliability): The degree of resemblance or consistency among repeated readings of the same thing. Simply put: How close the measurements are to each other. It is important to recognise that while a device might measure something with good precision by consistently and repeatedly reading the same value, the number may not reflect the true value, and therefore be inaccurate.
Precision does not mean accuracy! On average, CGMs are accurate but sometimes lack precision.
Applying concepts of accuracy and precision
Let's go back to the cycling example to help explain these terms a little more. Imagine cycling at your FTP and checking your chest strap heart rate monitor, which reads 129 bpm (identical to the ECG reading in the laboratory experiment from the week before). At the same time, your expensive pulse rate watch says 120 bpm. In this scenario, you can probably bet that your chest strap is more accurate than your watch.
However, if the chest strap is loose and moving all over the place, it may sometimes read 70 bpm or even nothing at all. You could now say that your chest strap is generally more accurate than your watch but sometimes lacking in precision. Buying a new chest strap that is a little snugger would allow you to have both a more accurate and more precise heart rate monitor!
But what if you still want to use that expensive watch...
If your watch always reads 5 bpm lower than the chest strap, you could say that your watch is rather precisebut not as accurate as the chest strap. Sometimes, being somewhat inaccurate but precise is okay, particularly if you know the device has a consistent "bias" (i.e., always reading a little low or high). You could mentally add 5 bpm to your watch readings, and then it's bang on!
While our example discussed a heart rate monitor, these same problems are often encountered with CGMs, which can sometimes be inaccurate and/or imprecise. In some instances, they can even be totally unreliable (like when they have missing data!). But, for the most part, newer CGMs are pretty darn good, at least in carefully controlled studies where frequent blood sampling is used as the gold standard comparator (1). In these studies, tons of data and several sophisticated statistical tools were used to describe CGM accuracy in manners such as Bland-Altman plots, 15/15% sensor agreement analysis, Clarke error grid analysis and mean absolute relative difference (MARD). In one recent accuracy study of the FreeStyle Libre2TM in healthy women, the CGM sensor was deemed pretty accurate, with CGM readings within 15% of the true value (as measured in the venous circulation) roughly half the time. But, the CGM did appear to read a little higher, on average, when the glucose levels were low (i.e., <4 mmol/L or < 72 mg/dL), and when glucose levels were changing rapidly, accuracy was found to be a little worse.
Understanding CGM delay
However, one could contest that these accuracy studies are not ideal when testing CGM accuracy since CGMs measure glucose in the interstitial fluid and not in the bloodstream. Glucose flows first into the bloodstream and then into the interstitial space, with a time delay of about 10-20 minutes (see the infographic below).
So, when glucose levels are changing rapidly, like after a meal or during heavy exercise, CGMs appear to be less accurate. However, this is simply a delay issue and NOT an accuracy problem (2). This delay is depicted in the image below, where the CGM accurately reflects what is occurring in the bloodstream but with a delay. If you measure glucose levels in the bloodstream and interstitial fluid at the same time, it looks like the CGM is inaccurate. But it’s not. It is accurately measuring the glucose in the interstitial space. That’s its job!
Furthermore, the lag time difference between blood and interstitial fluid may be up to 20 minutes during exercise, but the CGM eventually catches up when a steady state occurs. So, if you have a glucose meter, it is important to consider this time delay before accusing the CGM of being inaccurate. If you want to compare a blood glucose meter to a CGM value, it's best to compare it when glucose appears flat and stable. Then, as long as your hands are super clean when you poke your finger (don't have any energy Gu or gel on them!) and your blood glucose meter is accurate, you can compare your glucose meter to the CGM readings, and even do a CGM calibration if it allows for that!
CGM improvements over time
Earlier CGM devices were less accurate than they are today and were often unreliable. They only lasted for 3 days, and the values could be way off. They sometimes gave no value for several minutes, and the glucose values occasionally appeared suspiciously jumpy or seemed unbelievable (for instance, if you just ate a huge sweet treat, the CGM values did not budge at all or somehow dropped). To help with this, the manufacturers designed CGM systems that could be "calibrated" to a medical device called a blood glucose meter. Those meters were considered accurate, but in fact, some were NOT. This brings up another issue (which we need more time to delve deeper into) related to calibration errors with an inaccurate so-called gold standard.
Since the development of the first CGMs, companies have made significant strides in enhancing the safety and accuracy of their CGM devices. If, for example, a CGM reads suspiciously low or high compared to its recent values (i.e., the signal is jumpy or noisy), the system will not display any value. This was a safety measure for people living with diabetes since these values can influence how they take their insulin or glucose medication. Newer CGM systems are considerably more accurate and reliable as they typically come factory-calibrated since not everyone has a good quality (i.e., accurate) blood glucose meter for calibration purposes (and why poke your finger if you don't have to!).
All commercial CGM units available today in Europe and North America must be tested extensively for accuracy before being sold. Accuracy studies of most CGM units show that glucose readings are typically within ~10% of the true value. For example, if your blood glucose is actually 100 mg/dL, then a CGM might display a value around 90 mg/dL or 120 mg/dL for your interstitial glucose (3).
However, these same studies show that from time to time, CGM values can be way off compared to glucose values in the bloodstream (blood glucose). This probably goes back to the aforementioned delay issue and the analogy/cartoon of the train (blood glucose) and caboose (interstitial glucose). Moreover, because glucose levels tend to rise and drop more quickly during exercise than at rest, CGMs are often thought to be slightly less accurate during exercise as compared to rest (4).
But again, this might not be exactly true. The CGM measures glucose concentration in the interstitial space, not the bloodstream. So, if the so-called gold standard is blood glucose, then the CGM will typically read a little lower when blood glucose levels are rising and a little higher when blood glucose levels are falling. This phenomenon is called a sensor lag. Nevertheless, they should be considered accurate with respect to what they are actually measuring, the interstitial glucose, not the blood glucose.
By the way... your muscles and central nervous system "see" interstitial glucose, not your blood glucose, so we feel interstitial glucose sensing may be more important than blood glucose for exercise performance. One study suggests that to help with CGM accuracy during exercise, wearing the sensor on the upper arm may be best (5).
Your muscles and central nervous system "see" interstitial glucose, not your blood glucose, so we feel interstitial glucose sensing may be more important than blood glucose for exercise performance.
Additional factors to consider
It is also pretty clear that, relative to your blood glucose levels, CGMs can display lower readings overnight, particularly if you are lying on your sensor. This is called a "compression" low. Furthermore, CGMs are also influenced by hydration status, and some devices can even give false readings with certain medications such as acetaminophen (e.g., Tylenol), salicylic acid (found on some topical medications) and high doses of vitamins like vitamin C (see a FreeStyle article here). Some systems have poorer accuracy at the beginning of wear time, while others may have a deterioration in accuracy on the last day of sensor wear. So, perhaps it is buyer beware for day 1 and day 15!
Context matters for CGM
For some purposes a more accurate instrument is needed, for other purposes the highest accuracy and precision would be overkill. Consider a timing device that can measure finish time to 1 thousands of a second. This may be important in an Olympic track cycling final, but in a local ultraendurance run, this would be overkill. For some research purposes the accuracy of a CGM may not allow you to answer all questions, but for almost all practical purposes (we will discuss these in the next blog), the accuracy and precision of a CGM is sufficient.
In summary, the newest CGM systems are generally accurate and reliable. But from time to time, the values may be completely off. This could be related to poor accuracy or (more likely) it might simply be a delay in glucose equilibrium between the bloodstream and the interstitial space where the glucose sensor is doing its measurements. So, we suggest not to freak out over one or two very high or very low CGM readings and instead try to focus on glucose trends over time.
Jin Z, Thackray AE, King JA, Deighton K, Davies MJ, Stensel DJ. Analytical Performance of the Factory-Calibrated Flash Glucose Monitoring System FreeStyle Libre2TM in Healthy Women. Sensors (Basel). 2023 Aug 25;23(17):7417. doi: 10.3390/s23177417. PMID: 37687871; PMCID: PMC10490447.
Zaharieva DP, Turksoy K, McGaugh SM, Pooni R, Vienneau T, Ly T, Riddell MC. Lag Time Remains with Newer Real-Time Continuous Glucose Monitoring Technology During Aerobic Exercise in Adults Living with Type 1 Diabetes. Diabetes Technol Ther. 2019 Jun;21(6):313-321.
Freckmann G, Eichenlaub M, Waldenmaier D, Pleus S, Wehrstedt S, Haug C, Witthauer L, Jendle J, Hinzmann R, Thomas A, Eriksson Boija E, Makris K, Diem P, Tran N, Klonoff DC, Nichols JH, Slingerland RJ. Clinical Performance Evaluation of Continuous Glucose Monitoring Systems: A Scoping Review and Recommendations for Reporting. J Diabetes Sci Technol. 2023 Aug 20:19322968231190941.
Moser O., Mader J.K., Tschakert G., Mueller A., Groeschl W., Pieber T.R., Koehler G., Messerschmidt J., Hofmann P. Accuracy of continuous glucose monitoring (CGM) during continuous and high-intensity interval exercise in patients with type 1 diabetes mellitus. Nutrients. 2016;8:489. doi: 10.3390/nu8080489.
Coates AM, Cohen JN, Burr JF. Investigating sensor location on the effectiveness of continuous glucose monitoring during exercise in a non-diabetic population. Eur J Sport Sci. 2023 Oct;23(10):2109-2117.
Disclaimer: Asker Jeukendrup is a consultant to Supersapiens. Michael Riddell serves as a scientific advisor to Supersapiens and as a consultant to Dexcom Inc, another CGM device company.