Bioequivalence and Bioavailability Forum

Hi all,


f2 has been tested vigorously and rigorously with fraudulent datasets, by ... well ... someone.
We can rather easily anchor it to some convenient value like zero for a perfect match and so on. No issues there. But even so, the experiments with f2 have been troubled by the fact that it does not separate duplicates from non-duplicates very well. So empirically speaking, it has been a kind of dead end.
However, that can be improved if we introduce weighting and give higher weights to higher concentrations just as with methods 101, 36, 205 etc.
Even so, a plain f2 derivative does not capture dilutions. So, if profiles are diluted before re-analysis then f2 can be a painful acquaintance. Moreover, the need to validate not just precision but also accuracy when developing methods is relevant: Even with un-diluted re-analysis the re-analysed profiles are occasionally displaced vertically. Stating it differently: We get dilution estimates a little bit different from 1 with methods that allow that, and that phenomon renders methods that do not allow dilution kind of useless. At least, so says the guy who has been looking a bit into it. He is a bit of a weirdo, though. I heard he is currently in Ahmedabad with trips to Surat, Mumbai, Goa and Hyderabad coming up.
Getting something useful from that guy might take a while.