22. August 2012

Meter Accuracy

Probably preaching to the choir with this post, but still:

I test twice with my meter every time I have a result >250. Sometimes the variations are huge, one example from last week: 428 mg/dl vs. 303 mg/dl. In this case, the deviation from the true value (which of course I don't know) has to be at least 17%. (Interested in the math? See below*.)

Meters in Germany, as well as in the US, need to have a deviation of no more than 20% (only ~60% of the approved meters on the German market even fulfill this standard).
Edit: Somebody pointed out to me, that this sentence is contradictory in itself. (Thank you, Holger.) So let me elaborate: Meters in the EU need to have an CE certificate, which officially requires the fulfillment of the ISO norm 15197:2003 (deviations <20% or <15mg/dl for BS <75 mg/dl). But this certificate is merely a statement by the manufacturer (or the importer), that the meter fulfills the standard European regulations. Nobody checks who checked for the fulfillment of the ISO norm (e.g. the manufacturer). The CE certificate can be issued by a lot of so-called Notified Bodies in the EU.
A study by the university of Ulm lead by Dr. Freckmann and published in 2009 showed that actually only 60% of the meters in Germany fulfill the ISO norm.
The manufacturers argument to legitimate low meter accuracy usually is, that with this deviation no decisions in your diabetes therapy change, and that for high blood sugars usual correction scales can't be used anymore anyway, due to changed metabolic processes. (Or at least that's what meter company representatives told my at an expo.)

Well, I still use my usual correction scale and it works fine, as long as I use the average of my two tests to calculate it! If I know that my high reading is due to a miscalculated meal or a skipped bolus, there is no reason for ketones or anything else which would give reason for a different correction scale. And even if there was, does any of you have one??? I certainly don't.

Using the normal correction scale on a reading that might be easily 60 points above my actual blood sugar, means that I will most likely end up with a reading 60 points below my aim. So at 40 mg/dl!

Dear pharmaceutical industry: A low like that can be life threatening, so don't tell me a 20% deviation does not lead to changes in therapy decisions! Even smaller deviations do!

With the tight control a lot of diabetics try to obtain, 15% or even 20% deviation are too much to fit our needs. I don't care whether my meter needs only small drops of blood, is quick and small, can be attached to an ipod, sends data to my pump, has background light, comes in pink or can sing songs: If it doesn't fit my needs of accuracy, it's not better than its predecessor! You are producing meters for 20 years now, and there is hardly any real progress in this area.

Also keep in mind, that the guaranteed deviation only has to be kept in 95% of the tests. So in the other 5% my values can be of by 30% or whatever.

And don't you dare shifting the blame to the patients for not washing their hands and whatnot! A lot of the patients are very thorough with their testing routine and it would be a shame not to improve meter accuracy because of the patients who wouldn't make use of it. Also, if your meters don't pass the on-road test that's not the patients fault but something you should put some serious R&D into.

As said before, I'm probably preaching to the choir here. But in my opinion it can't be stressed enough: Something has to happen in terms of meter accuracy, otherwise all of the other diabetes gizmos are useless!

15 years of developement and I don't see a difference.

Fine-tuning my pumps basal rates? Ha!

Hopefully I can follow-up this post with another one about meter accuracy soon, but I first need to find some additional information about the subjact-matter.

*If 428 is 12% above the true value, the true value is 428/112*100, because 428 are 112% of the original value. True value: 382
If 303 is 12% below the true value, the true value is 303/88*100, because 303 are 88% of the original value. True value: 344
As you see, the upper value is still above the lower value, after calculating in a 12% deviation. This means, the deviation has to be larger. As you can see in the small table below, I tried this for increasing deviations, and only at 17% deviation the upper value matches the lower value. This means, I might have had a blood sugar of 365mg/dl. 
With a deviation of 20%, my bs could have been between 357 and 379.

upper lower deviation
382     344    12%
379     348    13%
375     352    14%
372     356    15%
369     361    16%
366     365    17%
  ⁞          ⁞        ⁞
357     379    20%