(6172 d 05:11 ago)
Posting: # 169
you are raising a very interesting question!
Simple answer: Act according to the protocol!
But obviously such an event was not anticipated, so we have to go for a more complicated answer.
As a diving instructor
and cave diver
, I love this rule:
STOP — THINK — ACT
- If possible Stop your evaluation, do not assess BE in any way.
Regulators might get the impression you are on the slippery track of data mining, i.e., performing various evaluations and later on picking out the ‘best’ results.
- Now let’s Think. Obviously it’s not possible to calculate the elimination phase for this subject getting a reliable estimate of AUC∞.
According to e.g., European guidelines it should be possible to drop the subject, because the prerequisite of AUCt ≥80% of AUC∞ is not fulfilled.
On the other hand, dropping a subject would mean loosing statistical power due to the smaller sample size.
In order to maintain the maximal statistical power in your study, you may drop the subject from the evaluation of AUC, but keep him/her for the evaluation of Cmax.
Maybe your last time points are something like 48, 72, 96, 120, and 168 hours? You may present a supportive evaluation of AUC truncated at 48 hours (for all subjects). There are numerous studies (mainly by Midha, Endrényi, Tóthfálusi) demonstrating that the result of a BE assessment does not change for AUCs truncated at time points later than 2–4×tmax. The general accepted truncation time point at 72 hours is very conservative (coming from the highest observed residence time of an oral formulation in the GIT); 48 hours should do as well.
If your last sampling points were 72, 96, 120, 144, and 168 hours, you may even dare to define the new main target parameter for extent of exposure as AUC72 (and keep the subject in the assessment).
Given all these considerations it should be possible to plan a new confirmatory evaluation (it’s advisable to write an amendment to the statistical protocol, or at least a ‘Note to File’).
- Act according to your new procedure. It’s important to stick to one (and only one) main evaluation, otherwise you may run at the best into multiplicity issues, or at the worst raise suspicions of ‘data dredging’…
IMHO you should not encounter major difficulties with regulators.
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
The quality of responses received is directly proportional to the quality of the question asked. 🚮