CGM
CGM accuracy vs finger prick: learn what MARD scores mean, why sensors lag, and how to check your own readings with simple steps.
Table of Contents
- Key Takeaways
- The Short Answer
- What MARD Means and Why It Misleads
- How Accurate Are Finger Prick Meters?
- Why Company Studies Disagree
- Why Your CGM Is Always a Little Behind
- How to Check Your Own Sensor
- When to Trust Which Device
- Where the Evidence Is Still Missing
- Frequently Asked Questions
- References and Sources
Key Takeaways
- A finger prick measures sugar in your blood. A CGM measures sugar in the fluid between your cells. They are not reading the same thing at the same moment.
- Modern CGMs report a typical error (called MARD) of roughly 7.5 to 9.8 percent in company and review data. That means a reading of 100 mg/dL is off by about 8 on average.
- Finger prick meters must meet a legal standard: 95 percent of results within 15 mg/dL (below 100) or 15 percent (at 100 and above) of a lab result.
- MARD scores depend heavily on how a study is run. One company funded study put the same sensor at 8.2 percent, and another at 13.6 percent.
- Your CGM runs several minutes behind your blood. Studies put the delay between about 5 and 10 minutes, and it grows when sugar changes fast.
- The fairest way to compare two devices is when your sugar is steady, not just after eating.
The Short Answer
Neither device wins every time, because they answer different questions. A finger prick tells you what is in your blood right now. A CGM tells you where your sugar has been heading all day. When it comes to CGM accuracy, a good meter on paper usually lands closer to a lab number at a single moment, while a CGM gives you thousands of readings that no finger prick routine can match.
That answer is more useful than a single accuracy score, and it matters because many people buy a CGM expecting lab grade numbers on every reading. The real figures are good, but they come with limits you should know before you trust a spike or a dip on your phone.
What MARD Means and Why It Misleads
MARD stands for mean absolute relative difference. It is the average gap between your sensor reading and a reference value, shown as a percentage. A lower number means a closer match. A MARD of 8 percent means a sensor reading of 100 mg/dL is off by about 8 on average. At 200 mg/dL, the same sensor would miss by about 16.
Here are the figures reported in a published comparison of widely used CGMs, using adult data:
- FreeStyle Libre 2: 9.2 percent
- FreeStyle Libre 3: 7.5 percent
- Dexcom G6: 9.8 to 9.9 percent
- Dexcom G7: 8.2 to 9.1 percent
The Dexcom G7 range comes from its main US trial, where about 320 adults with diabetes wore sensors for 10.5 days at 12 sites. The arm sensors scored 8.2 percent and the belly sensors scored 9.1 percent. Other sources list the Libre 3 at 8.9 percent for adults, which Dexcom quotes from Abbott’s US user guide. So even one sensor can carry two published numbers.
Two cautions apply. First, these numbers come from studies in people with diabetes. Few trials report the same score for healthy people whose sugar stays between 70 and 140 mg/dL, so treat the figures as a guide for your own use, not a promise. Second, MARD is an average. A sensor with a good average can still miss badly on some readings, so the average hides the worst cases.
How Accurate Are Finger Prick Meters?
Home meters follow an international standard called ISO 15197:2013. It requires that at least 95 percent of results fall within 15 mg/dL of a lab reference when sugar is below 100 mg/dL, and within 15 percent when it is 100 or higher. At least 99 percent must also fall in the two safest zones of an error chart used to judge how a wrong reading could affect treatment.
Many meters beat that minimum. In a 2025 evaluation of one modern meter, the MARD was 4.15 percent. A review of meter data also found that the two measures are related but not the same. One strip lot with a MARD as low as 6.1 percent still failed the ISO test.
Technique matters as much as the meter. In a 2019 study published in Current Medical Research and Opinion, five systems were tested in the hands of ordinary users. Four met the ISO limits with at least 95 percent of results inside them. The most common user mistakes were not checking the strip expiry date, applying blood the wrong way, and handling the device incorrectly. Washing and drying your hands before a test is the simplest fix, because food on your fingers can raise a reading.
Why Company Studies Disagree
ment for every finger prick.In its own US trial, Dexcom reported 8.2 percent for the G7 on the arm. A later head to head study published in the Journal of Diabetes Science and Technology, funded by Abbott, reported 8.9 percent for the Libre 3 and 13.6 percent for the G7. The diabetes technology site Diabettech pointed out that the study was not registered on clinicaltrials.gov and that its G7 score sat far above Dexcom’s own figure.
You do not need to pick a side to learn something. The same sensor produced two very different scores depending on who ran the study, which people took part, and how the reference blood tests were done. Treat any single MARD number as one result from one setup. When two company funded studies disagree, wait for independent testing before you rank devices by a decimal point.
The US Food and Drug Administration sets its own bar for the CGMs it labels as integrated systems. According to a 2020 comparison table in Frontiers in Endocrinology, more than 87 percent of readings must fall within 20 percent of the reference across all readings, with extra limits for low, middle, and high ranges. That is a looser standard than the one meters must meet, and it is the reason a CGM is not a drop in replace
Why Your CGM Is Always a Little Behind
A CGM sensor sits just under the skin and reads the fluid between your cells, called interstitial fluid. Sugar has to travel from your blood into that fluid before the sensor can see it. That trip takes time.
Studies give different sizes of delay. A Mayo Clinic study of healthy adults who had fasted overnight found a travel time of about 5 to 6 minutes. A clinical study of 37 people measured a mean total delay of 9.5 minutes from blood to sensor reading, with the sensor’s own electronics and smoothing adding to the body’s share. Roche’s patient education page says the delay can stretch to 12 to 24 minutes when sugar changes fast, such as after a meal or a workout.
This explains a common puzzle. You eat a big plate of rice, check a finger prick at the 30 minute mark, and the meter reads higher than your CGM. Nothing is broken. Your blood moved first and your sensor is catching up. The same lag works in reverse as your sugar falls.
How to Check Your Own Sensor
You can run a simple check at home. Use a meter that you trust and fresh strips that have not expired.
- Wash and dry your hands. Then take your finger prick and read your sensor within a minute or two of each other.
- Test when your sugar is steady, such as before a meal or first thing in the morning, and not 30 minutes after eating.
- Work out the gap. Subtract the two numbers and divide by the meter reading. For example, a meter reading of 100 and a sensor reading of 108 is an 8 percent gap.
- Repeat on three or four different days. One test is a snapshot and several tests show a pattern.
A gap inside about 15 to 20 percent is in line with the limits that regulators use. Remember that your meter has its own error of up to 15 percent. That means two devices that both pass can still differ from each other by more than either one differs from a lab. If your gaps are often far beyond that on steady readings, contact the sensor maker for a replacement and ask your clinician about the readings you have.
When to Trust Which Device
Use your CGM for trends, patterns, and how foods and walks affect you across a day. Use a finger prick when the stakes are higher or the sensor looks off:
- You feel shaky, dizzy, or unwell and your sensor says everything is fine.
- Your sugar is changing quickly and you need to know the true blood value now.
- A reading looks impossible, such as a sudden drop after you slept on your arm.
If you take insulin or other diabetes medicine, follow the rules your care team gave you for dosing and for confirming sensor readings. An article is no substitute for those rules.
Where the Evidence Is Still Missing
The headline accuracy numbers for CGMs come from people with diabetes, tested across their whole range, including very high and very low sugar. The growing group of wearers without diabetes lives almost entirely between 70 and 140 mg/dL. Nobody has published a clean, independent head to head test of the popular sensors in that narrow window. Until that data exists, the best accuracy score for your sensor is the one you build yourself with repeated steady state checks.
Frequently Asked Questions
Is a CGM as accurate as a finger prick?
Close, but not identical. A good meter often has a lower average error at a single moment. A CGM has a higher average error but gives you continuous readings and trends that a meter cannot.
What is a good MARD score?
Lower is better. Current CGMs in published data sit between roughly 7.5 and 9.8 percent. Compare scores only from studies that used similar methods, because the way a study is run can move the number a lot.
Why does my CGM show a different number from my meter?
There are three common reasons. The CGM reads fluid between your cells and not blood. It runs several minutes behind your blood. And each device has its own allowed error, so two working devices can still disagree.
How long is the CGM delay?
Studies range from about 5 to 6 minutes in healthy, fasting adults to a mean of 9.5 minutes in a clinical study. Education material from Roche gives 12 to 24 minutes during fast changes.
Do I need to calibrate my CGM?
Not with the main current models. A published comparison table lists the FreeStyle Libre 2 and 3, the Dexcom G6, and the Dexcom G7 as factory calibrated. Check your own sensor’s guide to be sure.
Does sensor placement change accuracy?
It can. In the Dexcom G7 trial, the arm scored 8.2 percent and the belly scored 9.1 percent. Follow the placement your sensor’s maker approves.
References and Sources
- Accuracy and Safety of Dexcom G7 Continuous Glucose Monitoring in Adults with Diabetes. Diabetes Technology and Therapeutics, 2022. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9208857/
- Comparison of widely available CGM systems. Karger. https://karger.com/view-large/16428695
- Welcome to the Wild, Wild West: the CGM head to head study. Diabettech. https://www.diabettech.com/welcome-to-the-wild-wild-west-the-cgm-head-to-head-study/
- Dexcom provider site, G7 and FreeStyle Libre 3 accuracy figures. https://provider.dexcom.com/will-g6-products-still-be-shipped-patients
- Accuracy Evaluation of a New System for Self Monitoring of Blood Glucose With Three Test Strip Lots Based on ISO 15197:2013. https://pmc.ncbi.nlm.nih.gov/articles/PMC5851221/
- Mean Absolute Relative Difference of Blood Glucose Monitoring Systems and Relationship to ISO 15197. https://pmc.ncbi.nlm.nih.gov/articles/PMC9445334/
- System accuracy evaluation of the new blood glucose monitoring meter GLUCOCARD S onyx beyond ISO 15197:2013 requirements. Frontiers in Clinical Diabetes and Healthcare, 2025. https://www.frontiersin.org/journals/clinical-diabetes-and-healthcare/articles/10.3389/fcdhc.2025.1465732/epub
- Accuracy of five systems for self monitoring of blood glucose in the hands of adult lay users and professionals applying ISO 15197:2013 accuracy criteria and potential insulin dosing errors. Current Medical Research and Opinion, 2019. https://pubmed.ncbi.nlm.nih.gov/29927663/
- Accuracy requirements for blood glucose meters (ISO 15197:2013) and iCGM (US FDA), Table 7. Frontiers in Endocrinology, 2020. https://pmc.ncbi.nlm.nih.gov/articles/PMC7089921/table/T7
- Time Delay of CGM Sensors: Relevance, Causes, and Countermeasures. Journal of Diabetes Science and Technology. https://pubmed.ncbi.nlm.nih.gov/26243773/
- Time Lag of Glucose From Intravascular to Interstitial Compartment in Humans. Mayo Clinic study. https://pmc.ncbi.nlm.nih.gov/articles/PMC3837059/
- Lag time and Continuous Glucose Monitoring (CGM). Accu Chek. https://www.accu-chek.co.uk/training/cgm/lag-time
ABOUT THE AUTHOR .
ABIODUN EMMANUEL
An SEO content writer focused on creating clear, persuasive, and engaging content that educates and drives results.
About The Author
