Bioequivalence and Bioavailability Forum

Dear nvs!

❝ If the specification for the sampels in the run is, the IS response should

❝ be with in 50% of the average IS repsonse of the accepted calibration curve

❝ standards.

Interesting approach; did you follow one of Ohlbe's suggestions?

❝ The specification for the QCs in a run is, 50% of QCs at each level should

❝ pass the accuracy specification (+/- 15%)

OK, this point is the commonly accepted procedure according to FDA's guideline and the consensus papers of the 'Arlington Conferences I-III' (for references: I-II and III).

❝ 1) In a run one of the QC has failed in the accuracy acceptence i.e.,

❝ out of 15% accuracy, and the other is meeting the accuracy specification,

❝ but the IS response of the QC is not within the +/- 50% specification.

❝ Now, whether the run can be accepted, since one of the QC is meeting the

❝ specification. Or, since the passed QC is not meeting the IS acceptance

❝ limit, the run has to be declered as failed run?

IS variation is not a topic in the guideline. But since you have set a limit in your method, both QCs at the same level are rejected - and therefore the run is not valid.

❝ 2) What is the best practice for the IS variation specification.

❝ Means, considering the IS responses only from the accepted CCs, or from

❝ the aceepted CCs and all the QCs in a run.

First have a look at this thread.
Personally I still don't get the point, because the IS should protect the analyst against errors in volume transfers. If e.g., 50% of the solution are spilled in one of the sample preparation steps (after IS addition), the analyte/IS ratio will not change anyway.
Of course Ohlbe's argument about laking sensitivity close to the LLOQ are strong ones. One the other hand I have some doubts whether such a cut-off (50% of whatsoever) really would improve the credibility of our analyses.
Nobody would bother if e.g., a vial was broken at centrifugation and the sample simply is missing. AUC is pretty robust, and if the sample is close to t_max for sure a couple of others in the same section of the profile are available.
IMHO once you set a limit on the IS response - agreeing with Ohlbe - you must apply this rule for all samples (calibrators, QCs, and unknowns). Most likely you will have to face repeated runs - but what for?