bowell said:
Just using two meters will not give you any useful results :cry:
try testing with a test solution with a known BG level
The only way is to use three meters and use a voting system to vote out any one of the meters thats set % out of range from test solution .
if you only use two meters how do you know which one is correct ? if any
This is the way we analyze pp02 in CCR rebreather under water real life support equipment
Yes, I'm aware of the testing methodologies and the only reason I currently accept the OneTouch in preference to the Rx is because it agreed accurately with lab tests in the past. Of course, that might be coincidence given the stated accuracy of only +/-20%. The same would apply to three or more meters. Choosing arbitrarily between one or the other is, as you point out, pointless and I wouldn't do that.
Testing against a known test solution would be OK providing I could accurately produce one of my own but I don't have access to lab equipment. The supplied solutions only indicate an acceptable range of readings for the meters which are fairly wide. They don't give me a specific concentration so I can see a specific error. That's unfortunately the case with some of the published accuracy data as well - it just says x% of the readings were in the +/-20% range, so claiming 100% simply means no better than +/-20% which isn't very good. In any case, whole blood is quite different from a test solution which is why they only suggest a range: the haematocrit value can make a big difference as can a number of other factors.
A voting system wouldn't work in the case where the error range is greater than the spread of readings. It's a pragmatic solution which really only detects malfunction and is fine for checking equipment which has already been certified to a certain standard. In the case of these BG meters, that standard is so low that often the error range is greater than the differences we are interested in.
For example, if we tested with five meters, giving a range of 6.0 to 8.0, it's possible that the top two have a positive bias and should be discounted and so the true value ought to be much lower, perhaps 6.5. Or the reverse could be true, we just don't know without the data. Ruling out the extremes is making an assumption that might not fit the data. I've already seen some data that shows bias even when the spread of error is quite low.
In addition, there are some meters where the errors do not spread much at higher glucose levels, i.e. their accuracy doesn't decline much across the range of readings. Accu-Check meters seem to be good in this respect. Others seem to be much more variable.
It is these problems that motivate me to try and get some data from the manufacturers and try to assess them critically. The Rx guys have so far been very cooperative and are taking the questions seriously.