The FDA has posted their warning letter to Dexcom of March 4 on quality control issues with G6 and G7 sensors. It reveals some of the technical issues with verifying the accuracy of the sensors. There are no issues concerning sanitary conditions or other things relevant to users.
This letter had been mentioned in another thread as possibly contributing to the current shortage of G7 sensors in the US market.
Ooh, dynamite. I hadn’t been able to locate the actual FDA letter; thank you.
The FDA’s primary push seems to be that the construction of the G6 and G7 sensors was changed and this resulted in worse performance. So far as I could decode (the repeated use of (b)(4) did not help) the change to construction has been reversed in the G7 but there is some doubt about the the G6.
There are also lots of procedural complaints; basically tech-bro style “fake it 'til you make it” approaches to the required documentation.
But the dynamite is in the two (I think) references to AIDs and the one reference to failure to verify the start and end-of-life accuracy of the G7 (I think specifically, maybe the G6 too.)
What is going on?
[Your responses] do not address hazards and risks associated with the use of your device with automated insulin dosing (AID) systems which are commonly used with your G6 and G7 CGM devices.
We acknowledge your response states you are working with your AID pump partners to better inform this analysis and expect additional complaint data to inform this activity through March 2025 and plan for an update to the new AID Interoperability System Hazard Analysis, RA-1000096 by June 2025.
There’s a new document; a direct goggle search just links back to the letter. Looks like more government just round the corner.
The letter reads somewhat worse than that to me. It looks to me like the letter says that Dexcom changed the sensor (coatings and/or reagents) in a way that decreased the average accuracy and has been shipping them without first proving to the FDA that the new accuracy characteristics are safe and effective, especially for automated dosing. They also didn’t have proper monitoring and control over their test environment, so their testing of sensor batches might not actually have detected batches that were bad. The FDA was also concerned that their test reporting no longer distinguishes between batches of sensors where most were good but one was barely acceptable, vs batches where every single sensor was just barely acceptable. In the latter case, it’s no longer clear that the patient safety is as good as promised in the original approval. The FDA was also concerned about compression lows: having the Dex report LOW for extended periods could lead to hyperglycemia, and the safety of this hasn’t been proven in the submitted pre-approval clinical studies.
My take? First, Dexcom has played a bit fast and loose with the rules, perhaps because of the profit motive, and so the’re in a bit of trouble with the FDA. (The FDA actually threatened to pull all the sensors off of the market.) Second, those of us who have been using Dex sensors for a while know how far to trust the numbers, so the CGM actually does help us a lot. In practice we’re not getting into trouble with our closed-loop dosing algorithms, but someone not accustomed to CGM could properly feel tricked and lied to and outraged by the numbers that the CGM reports.
@bkh thank you for the very insightful post, coming from a lot better background than I had. It doesn’t reflect well on Dexcom, unfortunately. I have suffered from a fair number of those marginal sensors. Seems like they could afford to have more consideration for the customer now that the product is selling well.
TLAs run out PDQ; there are only 17576 of them. In this case the Google AI turns up only two; for “TLA” itself there are rather a lot!:
o “Artificial Insemination by Donor”.
o “Acute Infectious Disease”.
Of course diabetes is a non-acute non-infectious condition so I guess the second comes out as a bit ironic but the nice thing about AID is that it really is describing an aid! An aid for blood glucose control in PWDs.
One day we will have automated insulin delivery devices (AIDDs) which manage to deliver the insulin directly to our blood stream but until then all of us, T1 and T2 alike, should have the ability to get AID if we wish.
Yeah, my pluralization was wrong; it should have been to the QLA AIDD, AID is a mass noun. Why on earth have we made this stupid undesigned ambiguous language the official language of our country?
I know. I suppose another person might have described his comment about pluralization as “sarcastic” but the fact that I made the same mistake twice is ironic, to me. “AIDS” because of the Syndrome, so when people say “aids” they mean the QLA; AIDS, not the TLA “AID”.
This hadn’t struck me before. I don’t know where the TLA comes from; it should be AIS (if capitalised the way you did it) or ADS (Autoimmune Deficiency Syndrome); that is certainly the correct TLA.
Nope: the acronym is AIDS. It’s a QLA ; four letters (“quad” a word that everyone in America understands). My pluralization of “AID” was certainly incorrect, but your misunderstanding was too.