Is anybody tight with Insulet (Omnipod) and Dexcom? The two companies should get together and enable a free app to download that controls the insulin pump, while simultaneously reading CGM. Without it being stored anywhere but the "cloud", it surpasses HIPPA laws. That's how AgaMatrix squashed the issue with the iBGStar. That would be brilliant!
To add to the link I just posted with the caveat that this is all my own hypothesis.
The issue is not on the BG meter side but on the side of actually *controlling* delivery of insulin.
Take a look at what happened recently with Insulet and Abbott. Abbott produce a blood test system that meets FDA requirements - it's accurate to +/-20% 95% of the time *when tested*. Unfortunately at some point they managed to produce test strips that caused the FDA to worry; a recall occurred.
Some time later *after* the recall at least one Omnipod user reported on tudiabetes that her Omnipod was consistently reporting incorrect BG readings. Shortly afterward Abbott issued another recall specific to the Omnipod that required test strips that they had only just started to manufacture.
What matters here from our point of view is that the FDA issues generic standards for blood glucose meters and those only require that they read within 20% for 19 out of 20 tests *when tested*. The FDA standard does not distinguish between a BG meter that is consistently 20% high and one that delivers random results over a 40% range - they both pass with flying colors.
So far as I know this is not true of insulin pumps. An insulin pump has to deliver insulin with a given accuracy over time, so consistent errors are bad, random errors even out. I believe that accuracy is significantly greater than +/-20%; so if you ask for 5 units you don't get 4 or 6. I don't actually know because basic Googling doesn't give me any indication that there *is* a generic FDA standard for accuracy of insulin pumps.
The relevant consequence of this is that if a manufacturer connects a BG meter to an insulin pump the accuracy requirements change - they become the greater (accuracy) required of either system.
What is more if the FDA does not allow an insulin pump to *consistently* over or under delivery insulin by 20% a BG meter which (only) meets the FDA generic BG meter requirements can't be connected (directly) to the pump.
If, on the other hand, one of us sucker diabetics is put in the middle to transcribe the reading from the BG meter to the insulin pump, well, it's our fault if we get too much, or too little, insulin, isn't it?
That's just BG meters. When you start talking about CGMs, which don't, so far as I know, have to meet the FDA requirements for BG meters, it just gets worse.
What annoys me is that the approach the FDA takes to this is that Insulet can't make an Omnipod which is controllable by an Android app, even though I can trivially over or underdoes myself via the PDM. This is the nanny principle - I'm not allowed to control my PDM other than through an interface that the FDA approves because I might do something stupid unintentionally. But actually the FDA approved interface pretty much encourages stupidity (e.g. by double dipping on meal boluses).
Wow! John, than you for the insightful, in-depth, and intelligent response. It does infuriate me how lax the FDA is with a 20% margin of err for bG meters. If I get pulled for doing 20% above the speed limit, I'm going to site the Food and Drug Administration standards of diabetes glucose meters and say it was only 20% over the limit. It's cool, the federal government says so.
The 'carbohydrate' value permitted by the FDA on a food label only needs to be within 20% and for products where less that 90% of all the carbs comes from sugar the FDA only requires that the product contain *more* carbs than the lower limit.
Similar errors are permitted in the food databases and of course natural variations occur in all grown things (think how the sweetness of apples varies).
So the BG error permitted is comparable with that we expect to see when we measure food carbs. Add to the latter error our own errors estimating the weight of foods! For me that is another 20%, at least; more in a restaurant.
The issue I have with BG meters is solely connected to the fact that the FDA permits systematic errors (errors which are identical every time for a given meter/test strip lot) within the +/-20% limit. If the errors were random then, no problem; I have to deal with errors in excess of that anyway.
However, what we see in tests from a few years back is that moving to a new set of test strips will shift the readings. Very few systems were tested, but the tested systems all showed systematic variations between lots. The shifts were typically in the range 5% to 20%.
What this means is that we can be doing fine with a combo that reads high, and not doing any corrections, but then swapping to a new set of test strips that reads high suddenly starts to require corrections.
What has happened, of course, is that all the ratios programmed into the PDM become wrong, but detecting this is very difficult.
Nevertheless I regard this as a minor issue compared to the permitted, indeed mandated, underestimate of "total carbohydrate" in food labels.
 The FDA conformance requirements are here:
It is *not* easy to decode this document, but an accurate diabetic summary is "if you believe the food label all bets are off."
The methods used by scientists to measure carbohydrate in food leave a lot to be desired:
Once again it's difficult to understand this document but the summary is that you can expect the carbohydrate value to be the *least* accurate of the three major values (protein, fat, carbohydrate).
For a more readable story written by someone who sort-of got the point:
The simplifications used in the article understate the problems, particularly with regard to carbohydrates.
 Manufacturers must ensure that total carbohydrate on the label is at least 80% of the actual value. If the principle source of carbohydrate shows large natural variations to be sure that this happens the manufacturer must measure the *lowest* carb count found in the source (e.g. when it is least ripe etc) and, at the very most, give a "total carbohydrate" which is 25% higher than this. In practice manufacturers will probably simply quote total carbs as the lowest they ever see.