Public Workspaces

I can only comment from the perspective of the work in my own lab and in the independent labs whom I use; but the proposed data quality requirements seem rather loose to me.

6.2.5.1.1 R of 0.995 correponds to R2 of 0.990, which is rather low for chromatographic work (even with a manual injector). I personally prefer to see R2 > 0.995 and R > 0.997. But I guess it is ok. Not as egregious as some of the other proposals below.

6.2.5.1.2 Should single-level calibrations even be allowed? The better independent analytical labs that I know of use 3 to 5 point calibrations, and certainly the better manufacturer’s labs do the same.

6.2.5.1.3 A peak in the blank at 5% the response of the target analyte seems very large – this could lead to the analyte content being overstated by 5%, no? I personally prefer to see NMT 0.5% or even NMT 0.1%. Maybe that is not realistic for an independent lab's samples, but 5% seems too high to me.

6.2.5.1.4

a) Spike recovery of 70-130% may be ok for potency tests (however I would prefer to see 90-110% or at least 80-120%), but it is too tight for some impurities occurring at very low levels (ppm).

b) A % RSD of 25% between samples seems very large. I personally require NMT 5% or even 2%. Maybe that is not realistic for an independent lab's samples, but 25% seems much too high to me, at least for potency tests. (Might be ok for impurities occurring at low levels.)

6.2.5.1.5 A window of 80-120% seems much too large; and I can’t really understand the distinction between the 80-120% range in the first sentence and the 90-110% range in the second sentence. This implies it is ok for the calibration to be off by as much as 10-20% before the lab does anything about it? In my lab I typically see %RSD < 1% between all the standard runs, at least in HPLC work.

Now, in my lab I’m typically working with well-defined matrices and well-optimized methods; and even when I send samples to an independent lab, I am able to provide a lot of information. I’m sure many of the samples that an independent lab analyzes are more complex and not as well characterized, and certainly the methods can’t be optimized for every matrix. So maybe the criteria can’t be as tight as I would like to see.

However, I think the data quality requirements need to be tighter than what is in NSF's proposal, if the analytical results are to have any credibility at all.

More discussion!!

Good morning Ms. Eisner, Thank you for your comments on the Joint Committee ballot of Standard 173 (173i29) regarding the quality assurance for quantitative test methods. Attached are your comments on this issue and the response of the issue proponent, Kerri LeVanseler of NSF International. I hope the above addresses your concerns. Thank you for your comments. If you have additional comments, or wish to discuss these points further, please contact Ms. LeVanseler.