Continuous Glucose Monitors (CGM): Clinical Comparison

A data-driven breakdown of the major RT-CGM (Real-Time) and isCGM (Intermittently Scanned) systems available on the market.

Core Metrics Comparison

Data sourced from primary FDA pre-market approval documents and published peer-reviewed studies (references below). MARD (Mean Absolute Relative Difference) is the standard metric for accuracy; lower is better.

Device Overall MARD Wear Time Warmup Calibration Size (Volume)
Dexcom G7 8.2% 10 Days (+12hr grace) 30 minutes None required ~60% smaller than G6
FreeStyle Libre 3 7.9% 14 Days 60 minutes Factory calibrated Size of two stacked pennies
Dexcom G6 9.0% 10 Days 2 hours None required Requires separate transmitter
Medtronic Guardian 4 8.7% (with 780G) 7 Days 2 hours None required Rechargeable transmitter

Algorithm Integrations (AID)

A CGM's utility is magnified when paired with an Automated Insulin Delivery (AID) system. Current official integrations:

  • Dexcom G6: Tandem t:slim X2 (Control-IQ), Omnipod 5, Beta Bionics iLet
  • Dexcom G7: Tandem t:slim X2 (recently approved), Beta Bionics iLet
  • FreeStyle Libre 2/3: In development / partial rollout with Ypsomed and Omnipod 5 (Europe)
  • Guardian 4: Medtronic 780G (SmartGuard)

Common Mistakes in CGM Analysis

1. Overweighting MARD differences under 1%: The difference between 7.9% and 8.2% MARD is statistically significant in trials, but clinically negligible for most individuals due to insertion site variables.

2. Ignoring Vitamin C interference: Older generation sensors (like Libre 14-day) show false highs with Ascorbic Acid. G6/G7 and Libre 2/3 have largely mitigated this, but limits exist (e.g., >1000mg/day for G6).

FAQ

Does paracetamol/acetaminophen affect readings?

Dexcom G6 and G7 resist acetaminophen interference up to 1,000mg every 6 hours. Older Medtronic sensors may still show interference.

Why do first-day readings seem inaccurate?

The insertion trauma causes a local inflammatory response, leading to fluid shifts in the interstitial tissue. Algorithms filter this "sensor noise," but Day 1 MARD is universally higher than Days 2-10 across all devices.