Patients bring me their glucose graphs expecting a clean, exact number the way a blood draw gives one. A continuous glucose monitor doesn't work that way, and understanding why changes how you should actually use the data to shape meals, not just admire the trend line. The single most useful number for judging any sensor's reliability is its MARD score, and almost nobody explains what it means or where it can quietly fail you.
What MARD Actually Measures
MARD stands for Mean Absolute Relative Difference. It's the average percentage gap between what a CGM reports and what a lab-grade blood glucose analyzer reads at the same moment, across thousands of paired readings. A lower MARD means the sensor tracks closer to true blood glucose on average. It is not a promise about any single reading — it's a population average, so an individual number on your app at 3pm could still be off by more than the headline percentage suggests, especially at the low and high ends of the range where accuracy tends to be weakest.
How Today's CGMs Compare
The two devices most of my clients ask about are Dexcom's G7 and Abbott's FreeStyle Libre 3, and their published MARD figures are close enough that neither should be the deciding factor over cost, wear time, or which arm you don't mind wearing a sensor on for two weeks.
| Device | Reported MARD | Notes |
|---|---|---|
| Dexcom G7 (adults, arm) | 8.2% | Abdomen placement runs slightly higher at 9.1% |
| Dexcom G7 15-Day (adults 18+) | 8.0% | Marginal improvement over the standard G7 |
| Dexcom G7 (children) | 8.1% | Comparable to the adult figure |
| FreeStyle Libre 3 (overall) | 7.9% | First 14-day CGM to report a sub-8% overall MARD |
| FreeStyle Libre 3 (adults) | 7.6% | Best-case figure for the device |
| FreeStyle Libre 3 (children 6–17) | 8.7% | Slightly wider gap than the adult cohort |
Both sit in the same rough band, which is why I care far more with a client about what's pulling their specific readings off than about which brand's marketing sheet wins by half a point.
What Throws a Sensor Off
A handful of documented factors can push a CGM reading away from your actual blood glucose, and knowing them saves you from second-guessing a device that's actually working correctly, or trusting one that isn't.
- Compression lows. Sleeping on the sensor or wearing tight clothing over it reduces local fluid flow and produces a sudden, V-shaped dip in the graph that resolves once pressure is off. It looks alarming and it isn't a real low.
- Sensor warm-up. Most devices need roughly half an hour to two hours after insertion to stabilize. Readings during that window shouldn't drive a meal or dosing decision.
- Hydration status. Because CGMs sample interstitial fluid rather than blood directly, dehydration can limit that fluid and destabilize readings — one more reason fluid intake belongs in the same conversation as glucose management.
- Acetaminophen. High doses can chemically interfere with some sensors and produce falsely elevated readings, a well-documented interaction that predates the newest sensor generations.
- Vitamin C and aspirin. Excess ascorbic acid intake has been linked to falsely high readings on some systems, while salicylates in aspirin have been linked to falsely low ones — opposite directions, same lesson: supplement and medication timing matters.
- Physiological lag. Interstitial glucose trails blood glucose by several minutes, so a CGM will understate how fast a reading is moving right after a meal or a hard workout, even when it's functioning normally.
Where This Meets the Plate
None of this is a reason to distrust CGM data — it's a reason to read it correctly. When I'm using a client's trend data to adjust meal composition, timing, or portion size, I'm looking at the shape of the curve over days, not chasing a single number. A sensor with an 8% MARD is accurate enough to show you that a starch-heavy dinner spikes you higher than a protein-forward one, or that a walk after eating flattens the curve. It is not precise enough to be the sole basis for a clinical dosing decision during a compression low or the first hour after insertion.
This is also where pairing the graph with an actual food log earns its keep. A spike that looks identical on the screen can have two very different causes — a genuinely high-glycemic meal, or a sensor reading skewed by a nap spent lying on the arm it's attached to. Without the context of what was eaten, when, and how the sensor was worn that day, it's easy to draw the wrong conclusion from a graph that's technically working as designed. I ask clients to note meals, activity, and anything unusual — a new supplement, a long car ride, a hot afternoon — alongside the data, precisely because the documented accuracy factors above are common enough to show up in an ordinary week, not rare edge cases.
When a Fingerstick Meter Still Belongs in the Kit
A basic blood glucose meter isn't obsolete just because CGMs dominate the conversation. It remains the right tool to confirm a reading that looks physiologically implausible, to check a suspected compression low before treating it, or to verify accuracy during a sensor's first hours of wear. If a CGM number doesn't match how you feel — genuinely shaky, sweaty, or foggy at a reading that says you're in range — a fingerstick check is still the fastest way to settle the question, and it's a habit I encourage every client with a sensor to keep, not abandon.
The practical takeaway for choosing a monitor isn't chasing the lowest published MARD by a fraction of a percent. Both major systems now sit close enough together that wear time, sensor placement options, and how the app displays trend arrows matter just as much day to day. What matters clinically is knowing the specific situations — warm-up windows, compression, certain medications, dehydration — where the number on the screen deserves a second look before it drives a food or dosing decision.