Chromatography Forum

Hi,

I am wondering how are the values for %RSD and % recovery assigned when developing a method.

i.e in liquid chromatography assays, I often encounter methods with a 3-5% RSD and 2% standard recovery.

Would it be reasonable to assume this for GC assays as well?

Would it make sense for the %RSD value to be higher than the % recovery? i.e 3% RSD and 2% standard recovery? Or vice-versa?

Any ideas or thoughts would be appreciated.

RSD is easier to deal with. Basically, the reproducibility of a measurement needs to be substantially better than the reproducibility of the product or process being measured. There is an entire sub-discipline of statistics called "gauge theory" that deals with these things but to a (very rough) first approximation, the RSD of the measurement should be less than about 1/3 the RSD of the product or process. So, for example, if you are analyzing a pharmaceutical tablet with an allowed variability of +/- 10% (RSD), you could reasonably tolerate and RSD of about 3% for the analysis.

Recovery is a more complicated issue. In principle, it can be almost anything; as long as it's consistent you can calculate around it. In that sense the RSD of the recovery is more important than the recovery itself, and that essentially is subsumed in the overall RSD of the analysis. All of that said, a method with 2% recovery would be unusable; I assume you mean 98% - 102% (which is fairly common).

As to why the values you cite are common, part of the reason is "because we can". 3% RSD is fairly easy to obtain for simple assays of major components (obviously more difficult for trace analysis / complex matrices / unstable analytes).

RSD is easier to deal with. Basically, the reproducibility of a measurement needs to be substantially better than the reproducibility of the product or process being measured. There is an entire sub-discipline of statistics called "gauge theory" that deals with these things but to a (very rough) first approximation, the RSD of the measurement should be less than about 1/3 the RSD of the product or process. So, for example, if you are analyzing a pharmaceutical tablet with an allowed variability of +/- 10% (RSD), you could reasonably tolerate and RSD of about 3% for the analysis.

Recovery is a more complicated issue. In principle, it can be almost anything; as long as it's consistent you can calculate around it. In that sense the RSD of the recovery is more important than the recovery itself, and that essentially is subsumed in the overall RSD of the analysis. All of that said, a method with 2% recovery would be unusable; I assume you mean 98% - 102% (which is fairly common).

As to why the values you cite are common, part of the reason is "because we can". 3% RSD is fairly easy to obtain for simple assays of major components (obviously more difficult for trace analysis / complex matrices / unstable analytes).

And then there is the Environmental world where numbers like this are a dream

Recovery is a more complicated issue. In principle, it can be almost anything; as long as it's consistent you can calculate around it. In that sense the RSD of the recovery is more important than the recovery itself, and that essentially is subsumed in the overall RSD of the analysis. All of that said, a method with 2% recovery would be unusable; I assume you mean 98% - 102% (which is fairly common).

As to why the values you cite are common, part of the reason is "because we can". 3% RSD is fairly easy to obtain for simple assays of major components (obviously more difficult for trace analysis / complex matrices / unstable analytes).

I am a little confused about the part about "unusable" and how recovery can almost be anything.

Lets say for example, a procedure indicates the %RSD of 6 standards injections to be NMT 3.0%, and the standard recovery of the standard injection after every 10 injections is within 2% (98.0%-102.0%). I would think that it would not be reasonable for the recovery to be, let's say within 50% but how does one come to the conclusion that 2% is reasonable versus, let's say 5%?

Behind at least half of those commonly used limits is a compendium (if not several).
ASTM, USP, AAOC, EP, JP...

For example, USP generally allows 2% RSD for n=5 and requires at least 6 standards if you want to allow a higher %RSD (which you typically do for things that are not assays of some sort).

Don't get hung up on the recovery thing - I suspect you meant a 2% difference is allowed between standard and check standard response while Tom addressed the same thing as a 98.0-102.0% recovery.

These same compendial limits are typically quite a bit looser for GC unless internal standards are in use.

If you like to work with much wider limits, work on biological assays, not small molecule chromatography.

Behind all of these RSD requirements and such is supposed to be a sound validation report.

I would think that it would not be reasonable for the recovery to be, let's say within 50% but how does one come to the conclusion that 2% is reasonable versus, let's say 5%?

A recovery value of 50% would require a *lot* of explaining and justification as to why you can't do better, but if that's the best you can do and if the recovery is *always* 50%, then you simply factor that in to the report. Any variation in the recovery would then be reflected in the overall variability of the method.

Again, there are only two real criteria:
- how precisely do I need to know the results?
- if I can't get there, how precise can I get?