How to determine quantitation limit (QL)?

[LCJEN]

Hi
I work in a pharmaceutical company, as a general practice in the company, QL is determine by inject QL solution 6 times, RSD% must NMT 10% and accuracy @ QL must met NLT 80% and NMT 120%.
My problem is I don’t have a sample that is clean enough (without the analyt of interest) to do the accuracy spiking. The area count of the sample was ~60, and the QL solution area count was ~6, the recovery was low. I had to spike at high level than QL sol in order to meet recovery NLT80% and NMT120%.
What should I report as QL? 10ppm (met RSD% NMT10% but failed accuracy) or 100ppm (met both criteria’s)? I checked the ICH guideline, QL can be determined by signal to noise ratio, it has nothing to do with sample. Just wondering how other people in the field handle this problem? Thanks.

[lmh]

I personally would go, grudgingly, with the more pessimistic estimate.

A lot of how I felt would depend on whether my recovery is always low, or whether it just tends to have a large error because I am adding a small amount of material to a sample that already contains a lot (which is always a situation that makes for imprecise recoveries).

If the recovery is always low (the impression I get from your statement), then you are losing material. This means that you can't be sure that the ~60 that you already have is actually a true reflection on the amount of material already present. You may already be losing natural material from the sample (but because you don't have access to an analyte-free blank sample, and you don't know what the true answer should be, unfortunately you aren't in a position to know how much is being lost).

In wondering how to deal with issues like this, it's necessary to think carefully about what a limit of quantification means. You are claiming "give me X amount of material, and I can quantify it plus/minus an acceptable error". If the X that's already in the sample might be wrong, and our current best-guess of how wrong it is (by spiking) is unacceptably large, you can't claim to have met the requirements of a meaningful limit of quantification.

Another way to look at it: the low recovery issue would prevent you from measuring the native amount of material in the sample accurately by the method of standard addition, because the curve goes non-linear for very small additions.

If the recovery was low once, but high next time, and is basically just full of imprecision because you're trying to make small additions to a large amount of natural analyte in the sample, then there is hope that if you ever find a cleaner sample, your LOQ can be revised downwards. If the recovery is always low, it may be that your LOQ will never be improveable, even if you find a cleaner sample.

Good luck, it's a horrible situation to be in (though not uncommon!). It doesn't bear thinking about. For example, how do you prove specificity if you have no test material that doesn't give a signal?