Bioequivalence and Bioavailability Forum

Dear Shreyas,

Interesting question. Actually if you read M10, this criterion is mentioned for pre-study validation, but not in the acceptance criteria of in-study analytical runs, where only the back-calculated concentration of standards and QCs is considered.

However, I would not go into the direction you suggest:

• on a scientific point of view, if you have a signal at the retention time of your analyte and/or IS, it is likely to be due to a contamination or to carry-over. My preferred approach would definitely not be to raise the LLOQ so that it can be ignored and you can continue with a lousy method, faulty equipment or poor practices, but rather to identify and eliminate the cause. After all, you know the level of the interference in these 2 samples, but you can't be sure it is not higher in other samples;
  
• remember that if you reject the LLOQ sample and the second standard becomes the LLOQ, it retains its original acceptance criteria (± 15%). This may lead to unpleasant discussions with regulators and inspectors, on whether this limit is met despite the interference.

Hello shreyas goswami,

❝ If blank sample and zero standard have more that 20% peak area of LLOQ sample, can we consider second lowest standard as revised LLOQ for batch acceptance?

❝

❝ Can Sample Concentration below revised LLOQ be reanalyzed considering pass batch?

❝

❝ What can be the associated consequences other than analytical end?

I thought a bit about all this. I am reaching the same conclusion as the other respondent, but for different reasons.

1. First of all, in analytical chemistry there is a semi-fancy approach called "standard addition" which is basically just a kind of mathematical trick for quantitation in matrices under circumstances when a zero sample can't be produced. I think you have inadvertently stumbled on a scenario where your setup kind of fits in the standard addition way of thinking, but it has no particular applications in BE (yet, but how about K+, anyone?!?).
2. If your matrix contains a bit of analyte -I am not saying it does or it doesn't as I can't tell from your description- then once your ordinary regression is done you might have a method which has a small error at high concentrations and a larger error at lower concentrations.
In BE we are interested in the ration of T/R so, if lower concs are more affected than larger concs, and if the true ratio differs from 1, then one of the products will likely have a higher error / bias in its AUC than the other. The estimated AUC ratio will be off. Potentially also Cmax.
3. You have no way of arguing that the regrrssion gives you the right concentrations - you are reporting concentrations with an unknown error (and this part could be unrelated to what you get in terms of accuracy etc).
4. Most CROs that I know of use 1/x or 1/x² weighting. Lower concentrations weigh heavier for the regression than high concs. If the metrix is not analyte-free then the lower concs, as stated above, have a higher degree of relative difference between true and planned analyte levels, so the regression model will somehow favour those samples that have a larger relative difference between which again means you do not quite know if the reported concs correspond to the true concs regardless of QC accuracies.
5. If you get a signal with your current matrix and there was none during method validation, then your current setup can be argued not to correspond the validation circumstances.

I see potential trouble ahead. Why not just stop here and fix what's broken?