Last reviewed August 21, 2026 by the Medically Modern coverage team. Figures come from the peer-reviewed clearance studies and Dexcom’s US product documentation, both cited at the end.
Accuracy in continuous glucose monitoring is usually reported as MARD — mean absolute relative difference. It is the average gap between what the sensor says and what a laboratory analyzer says at the same moment, as a percentage.
A MARD of 8% means readings are off by roughly 8% on average. At a glucose of 100 mg/dL that is about 8 mg/dL. At 250 mg/dL it is about 20 mg/dL, because the figure is a percentage rather than a fixed amount.
The number most people should care about is not MARD. MARD is an average, and averages hide their outliers. A device could post a good MARD while still producing the occasional badly wrong reading. Agreement rates — how often readings land within a set distance of the reference — describe real-world behavior better, and they are in the table below.
The clinical trial numbers
Two studies matter, one for each sensor.
The Dexcom G7 study enrolled 316 adults with type 1 or type 2 diabetes across 12 US sites, wearing 619 sensors over 10.5 days and producing 77,774 sensor-to-reference comparisons.
The Dexcom G7 15 Day study enrolled 130 adults and produced 20,310 comparisons.
| G7, arm | G7, abdomen | G7 15 Day | |
|---|---|---|---|
| Overall MARD | 8.2% | — | 8.0% |
| Within 15% | 89.6% | 85.5% | 87.7% |
| Within 20% | 95.3% | 93.2% | 94.2% |
| Within 30% | 98.8% | 98.1% | 98.9% |
| Within 40% | — | — | 99.8% |
Read the “within 20%” row as the practical one. For an arm-placed G7, 95.3% of readings landed within 20% of the laboratory value. The remaining 4.7% did not — roughly one reading in 21. Across 288 readings a day, that is not a rare event, and it is the honest reason to keep a meter in the house.
Dexcom described the G7 at clearance as the most accurate CGM cleared in the US. That is Dexcom’s claim about its own device; treat it as a manufacturer statement rather than an independent finding.
Arm placement measured better than abdomen
The G7 study reported arm and abdomen separately, and the arm won on every measure — 89.6% versus 85.5% within 15%, and 95.3% versus 93.2% within 20%.
The gap is not enormous, but it is consistent, and it is the reason the back of the upper arm is the default recommendation. The abdomen remains approved for adults, and it is a reasonable choice for someone who cannot use their arms. Just know you are trading a little accuracy for it. Site options by age are on the sensor page.
Five things that push readings off
1. Pressure on the sensor
This is the most common cause of a reading that looks alarming and is not. Lying on the sensor compresses the tissue around the filament and restricts local blood flow, so the glucose in that small pocket of fluid genuinely drops — while your actual blood glucose has not moved at all. The result is a compression low.
The signature is recognizable: a steep drop with no cause, usually while asleep, that recovers within minutes of rolling over, and no symptoms of hypoglycemia. A fingerstick taken during the episode reads normal.
The fix is placement. If you sleep on your left side, wear the sensor on your right arm.
2. The first hours of a new sensor
Insertion causes a small amount of local trauma, and the filament needs time to settle. Readings in the first hours after warm-up tend to be the least settled of the whole wear. If a brand-new sensor looks off, give it a few hours before concluding it is faulty.
3. Glucose changing fast
The G7 measures interstitial fluid, not blood, and interstitial glucose trails blood glucose by several minutes. When your glucose is moving quickly — after a meal, during exercise, as a correction dose takes hold — the sensor is reporting where you were a few minutes ago.
This is physiology, not a fault, and no CGM avoids it. It means a sensor and a meter can disagree at a moment when both are working correctly. It also means the trend arrow is more informative than the number during rapid change.
4. Two medications
- Hydroxyurea. Readings run falsely high, which can mask a low. Dexcom’s instruction is explicit: do not use the G7 for diabetes treatment decisions while taking hydroxyurea, and talk to your doctor about another way to monitor.
- Acetaminophen above the maximum dose. A standard or maximum dose — 1 gram (1,000 mg) every 6 hours — is fine, and Dexcom states readings can still be used for treatment decisions at that level. Above it, readings may read high.
Dexcom reports no known interference from ibuprofen. If you are on a long-term medication and unsure, ask your pharmacist rather than guessing.
5. Scar tissue from reusing the same spot
Applying sensors to the same small patch of skin builds up scar tissue, and scarred tissue is poorly perfused, which degrades readings over months. Rotating sites is not fussiness — it protects the accuracy you are paying for.
Want to know what a G7 would cost you?
We verify your benefits with your insurer and give you the monthly figure before you commit.
When to reach for a fingerstick
The G7 needs no routine calibration and no scanning, so it does not require the daily fingersticks an older meter would. It does not remove the need for a meter entirely.
Dexcom’s own guidance is to confirm with a fingerstick when your symptoms do not match the reading. In practice:
- You feel low and the sensor says you are fine, or the reverse.
- A reading changes drastically with no explanation, especially overnight.
- You are about to treat a low or take a correction dose and the number looks implausible.
- The sensor is new and behaving unexpectedly in its first hours.
Trust your symptoms over the screen. The device is a very good instrument, not an infallible one.
A working sensor can still look wrong
Most readings that look broken are explained by something on this page rather than by a defective sensor. Before requesting a replacement, check whether the sensor was compressed, whether it is in its first hours, whether glucose was moving fast, and whether a medication is involved.
Sensors do genuinely fail, though. In the 15 Day study, 73.9% of sensors lasted the full 15 days, so about one in four ended early. Dexcom replaces those, but the request has to come from you — the process is on the sensor page.
Frequently asked questions
Is a MARD of 8% good?
Yes. CGMs were considered clinically usable at around 10% MARD, and the leading current sensors sit between roughly 8% and 9%. The practical difference between an 8.0% and an 8.2% device is small enough that wear time, comfort, alerts, and cost should carry more weight in a decision than the decimal.
Is the G7 15 Day more accurate than the standard G7?
Its reported overall MARD is marginally lower — 8.0% against 8.2% — but the two figures come from separate studies with different participants and designs, so they are not a head-to-head comparison. Treating them as equivalent in accuracy is the safer reading of the data.
Why does my sensor disagree with my meter?
They measure different fluids. The meter measures blood; the sensor measures interstitial fluid, which lags by several minutes. A gap during rapid change is expected. A persistent gap while glucose is stable is more suggestive of a problem with the sensor or the site.
Does dehydration affect CGM readings?
It can. Interstitial fluid volume falls when you are dehydrated, which affects what the sensor is sampling. It is not among the interferences Dexcom formally documents, but staying hydrated is sensible regardless, and worth considering if readings drift during illness or hot weather.
Does accuracy get worse toward the end of a sensor's life?
The published studies report accuracy across the whole wear period rather than a day-by-day breakdown, so there is no clean public figure for the final day specifically. What Dexcom does publish is sensor survival: 73.9% of 15 Day sensors reached day 15. If readings become erratic near the end of a wear, treat it as a sensor nearing its limit and confirm with a meter.
Sources: Garg et al., “Accuracy and Safety of Dexcom G7 Continuous Glucose Monitoring in Adults with Diabetes” (PubMed), Garg et al., “Accuracy of the 15.5-Day G7 iCGM in Adults with Diabetes” (PubMed), Dexcom interfering medications, Dexcom G7 overview, Dexcom G7 FDA clearance announcement.