Bioequivalence and Bioavailability Forum

Hi Shuanghe,

❝ Thanks. That's a bit dark (maybe because too much scandals such as GVK/Semler/Synchron… ?)… but at least it makes some sense as to why some agencies may ask for it…

"Some agencies" is indeed the correct way to say it, it is not only ANVISA now.

Take a step back and look at it. CROs started out by testing R versus R to pass. Then regulators found out and started inspecting the sites to verify DA. How annoying. Then we had the switcharoo (credit to a well known WHO inspector for coining this term) which involved switching T and R for select subjects and reinjecting under false ID. It was extremely efficient in the sense it can make any study pass how ridiculously badly T matches R. Then someone published a paper about it. how ANNOYING. Initial disbelief, the author got angry emails saying that although the paper presented some interesting ideas, it was generally disconnected from real life. Only took a few months until the matter was re-confirmed by FDA's letters to two CROs. Basically, regulators pulled the carpet under CROs when they fiddled with the clinics (pharmacy) or bioanalysis to make a study pass.
So out of desperation, the bad CROs have turned their focus to data management (or would that be data mismanagement?) and pk/stats. This approach with TPDs works well, but it is difficult to plausibly rig a study that is wildly failing. So, it will be mainly applicable to studies failing "a little bit". Regulators are now starting to ask for analysis with scheduled time points. How annoying.

Dark? No.
Creative? Yes.

There is in fact one other reason which I can also mention: the author of the paper mentioned has worked with agencies to improve SaToWIB. One of the potential improvements was to make an objective function which is based on AUC match between any two PK profiles. So if you have two identical profiles (within the meaning of random analytical variability) they will have better matching AUCs if there is no difference in recorded time points. So if you use this kind of objective function you increase the chance of detecting reinjected profiles if you ignore the TPD and just go with the scheduled timings.
But linear regression and method 32 still works much better if you ask me.