Log in
Register|
nobody nothing 2018-02-27 20:02 (2410 d 02:03 ago) Posting: # 18472 Views: 3,957 |
|
|
Hy! Have a drug with "normal" half-life of some hours in the mean, but considerable variability (BOTH target params). So pilot will be huuuugh and fully replicate (planned: EU submission). Sponsor wants to make a little cheaper by having only 24 h sampling in the pilot, but considerably longer in the pivotal trial. Good idea? Regarding sample size estimation for the pivotal (including adjusted acceptance range for Cmax, but standard acceptance range for AUC...). Any thoughts or numbers?  Οὐδείς — Kindest regards, nobody |
|
ElMaestro ★★★ Denmark, 2018-02-27 22:42 (2409 d 23:23 ago) @ nobody Posting: # 18473 Views: 3,389 |
|
|
Hy yourself, ❝ Sponsor wants to make a little cheaper by having only 24 h sampling in the pilot, but considerably longer in the pivotal trial. ❝ ❝ Good idea? Regarding sample size estimation for the pivotal (including adjusted acceptance range for Cmax, but standard acceptance range for AUC...). That's likely going to be fine. It will regularly be Cmax that is the showstopper. Even if your AUC does not cover 80% and if you are not sampling to 72hrs, then the idea you get about AUC aspects will still be very qualified. Note one thing which is commonly misunderstood: When pilot trials have small sample sizes compared to pivotal studies (which is almost a definition) then they do not give good info about formulation matches in BE (or, where applicable, of superiority prospects, but this is often very dependent of placebo arms). Pilot BE studies are great for everything but screening for formulation matches. — Pass or fail! ElMaestro |
|
nobody nothing 2018-02-27 23:47 (2409 d 22:18 ago) @ ElMaestro Posting: # 18474 Views: 3,406 |
|
|
Hi Iceland II Sounds good. With highly variable Cmax point estimates are frequently jumping around like hell between studies, but with highly variable AUC the point estimate from pilot should be more reliable/less freakish (with a reasonably high n), right? Didn't have a look at the numbers recently... Thanks for reply! (-11.4°C right in front the window... still falling...) — Kindest regards, nobody |
|
ElMaestro ★★★ Denmark, 2018-02-27 23:58 (2409 d 22:08 ago) @ nobody Posting: # 18475 Views: 3,398 |
|
|
Hi Nobody, it is quite simple. Look at the pilot trial's CI for whatever you are interested in, such as the T/R. If the CI is for example 60-149 (and it might well be in a pilot. Even for AUC. Even if AUC's CV is not high) then you cannot really take any decision based on the location of T/R. Think about it. For two identical product, you will "quite likely" make the wrong decision whether your decision is. There was a weirdo who wrote a paper about pilot and repeat trials as development tools in BE. The author's only regret is that he presented the data the way he did. It would have appealed more to the broad audience if he had graphed the chance of taking the wrong decision in the various scenarios when the decision was to be based on the location of the apparent T/R. — Pass or fail! ElMaestro |
|
nobody nothing 2018-02-28 09:36 (2409 d 12:30 ago) @ ElMaestro Posting: # 18477 Views: 3,302 |
|
|
❝ There was a weirdo who wrote a paper about pilot and repeat trials as development tools in BE. The author's only regret is that he presented the data the way he did. It would have appealed more to the broad audience if he had graphed the chance of taking the wrong decision in the various scenarios when the decision was to be based on the location of the apparent T/R. Yep, next time you send me your draft and we discuss the data presentation.  GMR of AUC is of some interest to judge which formulation to take to the pivotal (one from pilot(s) or maybe something completely different?). Not so much for sample size. Would take 0.9 plus CVintra estimated for Test from pilot(s) and see where one ends (1 gazillion of HVs?)... — Kindest regards, nobody |
Ing. Helmut Schütz