Reasonable recoveries at the LOQ

[mohan_2008]

Our method usually gives an accuracy of 100% (approximately).

During the determination of LOQ (Quantitation limit) what would be a reasonable spec for the recovery.

Does it need to be 100% or less than 100% at the LOQ.

[Consumer Products Guy]

Why would you expect 100% recovery at the limit of quantitation? If you really need to assay at that level (and be near 100%), then doesn't LOQ = target ? We only check recovery at 80, 100, and 120% of target, as 80 and 120% of target are already quite a bit out of specifications.

[Suresh Seethapathy]

I normally look for recoveries between 80% and 120% at LOQ. I don't have a scientific basis for the numbers but have tried to follow the following document.

http://dnr.wi.gov/org/es/science/lc/OUT ... cument.pdf

[krickos]

Hi

I would like to "turn the argument around" abit.
Range (including lowest reporteble result) is typically determined on a range of concentrations where the precision, accuracy and linearity is accepteble.

With that in mind you might even do not have to bother to check accuracy at LOQ levels (defined based on various noise calculations) as other parameters might fail before you reach that level.

So rather set a criterion for accuracy that is accepteble for the intended range of the method, not LOQ itself.

In the pharma industry I only know of one agency the australian one that has a guide for accepteble recoveries based on target/impurity content:

http://www.apvma.gov.au/publications/gu ... ethods.pdf


Active/impurity content % Acceptable mean recovery
≥ 10 ----- 98 –102%
≥ 1 ------ 90 –110%
0.1 – 1------ 80 – 120%
< 0.1 ------ 75 – 125%

[Alex Buske]

[quote]
Active/impurity content % Acceptable mean recovery
≥ 10 ----- 98 –102%
≥ 1 ------ 90 –110%
0.1 – 1------ 80 – 120%
< 0.1 ------ 75 – 125%[/quote]

Thanks for the link. I think this going to be rather strange, as relative response factors should only be used if they are <0.8 or >1.2. so just from the rel.- response factor one gets a deviation up to 20% - which should affect the recovery if the experiment covers measurement as well as calibration.

Alex