The Dexcom G6 is one of the more accurate continuous glucose monitors available, with an overall accuracy rate (called MARD, or mean absolute relative difference) of about 9% in adults. In practical terms, that means if your actual blood glucose is 100 mg/dL, the G6 reading will typically fall between 91 and 109 mg/dL. It was the first CGM to receive the FDA’s integrated continuous glucose monitor (iCGM) designation, which requires the device to meet strict accuracy thresholds across different glucose ranges before it can be used for insulin dosing decisions without a confirmatory fingerstick.
What a 9% MARD Score Means in Practice
MARD is the standard way researchers measure CGM accuracy. It represents the average percentage difference between the sensor reading and a lab-quality blood glucose reference. A lower number means better accuracy. The G6’s 9% MARD in adults is a strong result, but it’s an average. Some individual readings will be closer to your true glucose, and some will be further off.
The sensor performs differently depending on your glucose range. At normal to moderately elevated levels (roughly 70 to 180 mg/dL), readings tend to be most reliable. At very low glucose levels (below 70 mg/dL) or very high levels (above 180 mg/dL), the margin of error can widen. The FDA’s iCGM classification requires the device to meet specific accuracy benchmarks at each of these ranges, including tighter standards at low glucose levels where an inaccurate reading could lead to a dangerous insulin dosing mistake.
Why Readings Sometimes Differ From Fingersticks
The G6 doesn’t measure glucose in your blood. It measures glucose in the interstitial fluid, the thin layer of fluid between your cells just under the skin. Glucose moves from your bloodstream into this fluid, but there’s a natural delay. This lag time can be up to 15 minutes, though it’s typically shorter. When your glucose is stable, the lag barely matters. When it’s changing quickly, like after a meal or during exercise, the sensor can trail behind what a fingerstick would show by a meaningful amount.
This is the single biggest reason people notice discrepancies between their G6 and a fingerstick meter. If you eat a fast-acting carbohydrate and check both devices five minutes later, the fingerstick will often read higher because it’s detecting the spike in real time. The G6 will catch up, but it takes a few minutes. The same happens on the way down: if your glucose is dropping fast, the G6 may still show a higher number than your blood glucose truly is.
The First 24 Hours Are Less Reliable
A new G6 sensor typically needs time to settle in after insertion. During the first day, accuracy tends to be noticeably lower than it is for the remaining nine days of the sensor’s life. One study in pediatric patients found that when the first day’s data was excluded, accuracy improved meaningfully, with the average error dropping to about 13.4% and 78% of readings falling within 20% of reference values. Many experienced users treat Day 1 readings with extra caution and rely more on fingersticks if they need to make critical dosing decisions during that window.
Acetaminophen and Other Interferences
Earlier Dexcom models were notorious for giving falsely high readings when users took acetaminophen (Tylenol). The G6 largely solved this problem at normal doses. FDA bench testing confirmed that standard doses of acetaminophen (up to 1,000 mg every 6 hours) cause less than 10 mg/dL of measurement error, which isn’t clinically significant.
Higher-than-recommended doses are a different story. At roughly two to three times the therapeutic blood concentration of acetaminophen, the sensor showed up to 22 mg/dL of positive bias, meaning it read higher than actual glucose. At even higher concentrations, the bias reached up to 70 mg/dL in a low-glucose scenario. So if you’re taking standard doses for a headache, your G6 readings remain trustworthy. If you’re taking large amounts, the sensor may report glucose levels that are artificially elevated.
Accuracy in Children
The G6 is approved for ages 2 and up. In pediatric settings, accuracy tends to be slightly lower than in adults, partly because children’s bodies are smaller and their glucose levels can swing more rapidly. Clinical data from pediatric patients (median age around 11 years) showed a MARD of about 13.4% after the sensor’s first day, with 78% of readings landing within the 20/20 accuracy standard. That’s respectable for a CGM in younger users, though it does mean parents and caregivers should expect a bit more variability compared to adult performance.
Factors That Affect Day-to-Day Accuracy
Beyond the inherent lag and first-day settling period, several real-world factors can push G6 readings further from your true glucose:
- Sensor placement: The G6 is FDA-cleared for the abdomen in adults and the upper buttocks in children ages 2 to 17. Placing it on other body sites (which some users do) can change how quickly glucose reaches the sensor and affect accuracy.
- Compression lows: Sleeping on the sensor or pressing against it can temporarily restrict blood flow to the area, causing a false low reading. If you see a sudden unexplained drop at night, rolling off the sensor often resolves it within minutes.
- Dehydration: When you’re dehydrated, interstitial fluid composition changes, which can make readings less reliable.
- Sensor age: Accuracy is generally best in the middle days of a sensor session (roughly days 3 through 7) and can drift slightly toward the end of the 10-day wear period.
How It Compares to Fingerstick Testing
Home blood glucose meters themselves aren’t perfectly accurate. The FDA allows meters to be within 15% of a lab reference value for readings above 75 mg/dL, and within 15 mg/dL for readings below that. So when you compare your G6 to a fingerstick and see a 10 to 15% difference, both devices may be within their normal accuracy range. The “true” number could be somewhere between them.
Where the G6 has a clear advantage over fingersticks is in trend data. A single fingerstick tells you where your glucose is right now. The G6 gives you a reading every five minutes, 288 times per day, with trend arrows showing which direction you’re heading and how fast. For catching overnight lows, post-meal spikes, and gradual patterns over weeks, that volume of data is far more useful than a few daily fingersticks, even if any individual G6 reading is slightly less precise than a well-performed fingerstick.
When to Trust the Number and When to Double-Check
For routine diabetes management, the G6 is accurate enough that most users can dose insulin based on its readings without a fingerstick confirmation. That’s exactly what the FDA’s iCGM designation allows. However, there are situations where a fingerstick backup is wise: when the sensor reading doesn’t match how you feel, during the first 24 hours of a new sensor, when your glucose is changing rapidly, or when you’re getting a reading that would lead to a large insulin correction. If the G6 says you’re 55 mg/dL but you feel perfectly fine, a fingerstick is a reasonable next step before treating a low that may not be real.