Bioequivalence and Bioavailability Forum

Dear Helmut,

❝ I'm just trying to read tea leaves - or in other words, regulator's minds.

Good luck

OK, regarding your question: I think the idea is that a validation performed with runs of maximum 30-40 samples may not predict what will happen with longer runs. This was mentioned several times by various regulators in presentations (CT Viswanathan had that in a presentation at the BSAT in November 2005, if I'm not mistaken). The difference in run size between method validation and study phase was identified as one of the reasons why shit eventually happened during study sample analysis but was not identified during the validation. I'm sure it must be also in a white paper somewhere but I'm afraid I can't find it right now.

In your case I wouldn't be too worried unless you had a large number of rejected runs in your previous studies: you have tons of historical data to show that your method performs correctly also with full-size runs. I'm sure your QC samples were nicely placed within the runs to show that there was no drift or trend of any kind. Maybe you could compile these data and build a document with nice figures, to be appended to the validation report in the final report of any new trial, to justify why you don't need to do it again.

❝ Let's say we are able to inject 25 times from a single vial; would you say that it's possible to extract only four blank samples → four vials → 25 injections / vial → 100 chromatograms or extract 100 blank samples → 100 vials → chromatograms?

Not blank samples: 3 levels of QCs (or rather 4 with LLOQ ?).

I don't think this would be OK. Problems do not come only from the injector/column/detector, but may also come from sample handling and extraction, depending on the extraction method. It takes more time to extract 120 samples than a handful. In addition I suppose you would also be expected to report precision, though this is not clear in the draft (according to line 217 precision should be calculated using the data obtained for accuracy. Hopefully this will be clarified in the final version). Multiple injections of a single extract would not provide relevant information on the precision of the method.

Regards
Ohlbe