I switched from the Minimed Guardian to the Dexcom G6.
I hated all the calibration alarms waking me up through the night, feeling tethered by a “ball and chain” and pump failures at the absolute worst possible times.
That said, all the calibration alarms and hassle kept the sensor very accurate but the Guardian lowest basal closed loop BG set point was much higher than my body likes to ride at.
I moved to the Dexcom G6 because I wanted to have peace of mind to sleep through the night, predict and prevent lows, and less of a ball and chain than The MiniMed system, and to try the DIY Loop and set my BG basal set point lower.
The conclusion I’ve come to since switching to the Dexcom G6 sensors is that it is unpredictable and wildly unreliable. I cannot trust the Dexcom G6 to base treatment decisions on and definitely cannot trust the sensors to used in a semi closed loop insulin pump.
The pros is that it “can” help indicate and prevent hypothermic reactions.
My first box of 3 sensors went something like this:
Sensor 1: ~97% accurate
Sensor 2: ~80 accurate
Sensor 3: ~40% accurate or worse
I define accuracy as “Any reading, within a hour period, that will cause a correction bolus to be administered or withheld that would result in BG being out-of-range”.
so, if anytime within an hour I get a reading that would falsely indicate a correction bolus, or lack of, that would’ve resulted in a hypoglycemic or hyperglycemia reaction the pump has failed that hour.
I’d like to expand on this definition in the future to include a severity metric. Like “any correction bonus that would result in BGs being more than 40mg/dl
Sensor 3 was a nightmare. Everyday it wanted multiple calibrations but still could never hold an accurate reading. The best I could get was 30/30 and then an hour later it would be 60-80 mg/dl off again and I’d be fighting to re-calibrate.
So, I’ll simply write sensor 3 off as a failed sensor and not consider it in my overall assessment of the G6.
That leaves us with 2 sensors, 97% and 80%. Making treatment decisions that are life and death with inaccurate data 3-20% of the time is a problem.