Bioequivalence and Bioavailability Forum • Accuracy versus Inaccuracy

NewInPK
☆

2009-06-11 04:42
(6224 d 02:00 ago)

Posting: # 3843
Views: 6,305

Accuracy versus Inaccuracy [Bioanalytics]

Hello everyone,

I've used RE(%) = Accuracy for a long time. However lately I am debating if I should join the ranks of those who use RE(%) = Inaccuracy.

RE(%) = 100*(measured - theoretical)/theoretical

In many math books Accuracy is described as RE, which means that the lower the RE, the better the accuracy. However in day to day language, I'd rather be 100% accurate, then 1% accurate. In one place I noticed the term "Relative Accuracy Error", which mitigates my dilemma a little better.

Should we use Inaccuracy, Accuracy, or Relative Accuracy Error?

Same goes for precision.

Also, is %Bias = %RE?

I put this in the Off Topic area, as it is not that important.

Thank you for your input.

Edit: Moved to the Bioanalytics Category, because it is important. [Helmut]

Helmut
★★★

Vienna, Austria,
2009-06-11 06:19
(6224 d 00:22 ago)

@ NewInPK
Posting: # 3844
Views: 5,356

Accuracy versus Inaccuracy

Post reply

Dear NewInPK!

❝ I've used RE(%) = Accuracy for a long time. However lately I am debating

❝ if I should join the ranks of those who use RE(%) = Inaccuracy.

❝

❝ RE(%) = 100*(measured - theoretical)/theoretical

Welcome to the club!

❝ In many math books Accuracy is described as RE, which means that the lower

❝ the RE, the better the accuracy. However in day to day language, I'd rather

❝ be 100% accurate, then 1% accurate.

Yes!

❝ In one place I noticed the term "Relative Accuracy Error", which

❝ mitigates my dilemma a little better.

Hhm, I'm not a native English speaker - sounds a little bit weird to me. I must confess that I used both Accuracy and Inaccuracy in conversation - it was always clear from the context, which term I actually meant. In my reports I try to use Inaccuracy.

❝ Should we use Inaccuracy, Accuracy, or Relative Accuracy Error?

I would say:

Accuracy = 100*measured/theoretical, whereas
Inaccuracy = 100*(measured - theoretical)/theoretical

The Arlington III Whitepaper states for LBAs:

For a method to be considered acceptable, it is recommended that both the interbatch imprecision (%CV) and the accuracy, expressed as absolute mean bias (%RE) be ±20% (25% at LLOQ and ULOQ). As an additional constraint to control method error, it is recommended that the target total error (sum of the absolute value of the %RE [accuracy] and precision [%CV] be less than ±30% [±40% at the LLOQ and ULOQ]).

❝ Also, is %Bias = %RE?

IMHO, yes.

❝ Same goes for precision.

That's nasty. Arlington III again:

[...] the major sources of variability [are] imprecision and inaccuracy.

To quote William S. Burroughs:

Language is a virus from outer space!

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

NewInPK
☆

2009-06-12 06:16
(6223 d 00:26 ago)

@ Helmut
Posting: # 3849
Views: 5,299

Accuracy versus Inaccuracy

Post reply

Good evening HS,

Thank you very much for the link and for taking the time to reply.

I feel I am in very good company when it comes to interchanging these terms:

"For a method to be considered acceptable, it is recommended that both the interbatch imprecision (%CV) and the accuracy, expressed as absolute mean bias (%RE) be ± 20% (25% at LLOQ and ULOQ)."

Then the very next phrase:

"As an additional constraint to control method error, it is recommended that the target total error (sum of the absolute value of the %RE [accuracy] and precision [%CV] be less than ± 30% [ ± 40% at the LLOQ and ULOQ])."

I guess we accept easily that the accuracy of a measurement of 100 mL is ± 1 mL (also absolute error), but when it is relative and it becomes 1%, then it doesn't sound that reassuring, so there comes "inaccuracy" to help. I may have to use only "%RE" to avoid comments from the two camps.

Thanks again!