Bioequivalence and Bioavailability Forum

Dear Helmut,

❝ Let’s explore the example from above (x = 2–200, n = 10).

In my experience most labs would only have 8 calibration levels, not 10. And a 200-fold difference between LLOQ and ULOQ would be quite common in BE (possibly wider in pre-clinical). But OK, the principle remains the same.

❝ I would place [...] the HQC maybe between St9 & 10 (170 = 85% of ULOQ).

85 % of the ULOQ may be a bit high. You are likely to get some values above the ULOQ. Not a problem during study sample analysis (the QC fails, that's all), but more difficult to handle during method validation, where you have to calculate precision and accuracy.

❝ There is no gap in calibrators.

Not on a log scale... But looking at it on a linear scale you have no calibrator between 120 and 200, which represents a significant portion of your range.

❝ ❝ (where does it become non-linear ?)

❝

❝ Nowhere; it’s nonlinear in the entire range.

Sure. But looking at the calibration plots the first portion of the curve often looks quite linear and would actually perform quite well with a linear fit. It is only the highest calibration samples that justify a quadratic fit. If I take your example, what I have seen a few times is situations where looking only at the calibrators from 2 to 120 I would go for a linear fit. Then the 200 calibrator has a lower response. What is it due to: a really non-linear response ? Or just a spiking issue ? With no point in between it becomes quite difficult to answer that question.