Shah
☆    

India,
2021-07-16 15:40
(986 d 08:40 ago)

Posting: # 22472
Views: 2,729
 

 HVD Pilot: 3 candidates, 1 reference [Design Issues]

Can anyone please suggest the study design for pilot study with three test formulations and one reference formulation for highly variable product.


Edit: Subject line changed; see also this post #2. Please follow the Forum’s Policy[Helmut]
ElMaestro
★★★

Denmark,
2021-07-16 15:51
(986 d 08:28 ago)

@ Shah
Posting: # 22473
Views: 2,185
 

 HVD Pilot: 3 candidates, 1 reference

Hi Shah,

❝ Can anyone please suggest the study design for pilot study with three test formulations and one reference formulation for highly variable product.


More info about the purpose would be great. But in the absence of any more detail, I would go for a 5-period, 4- treatment, n-sequence BE design, where n (the number of sequences, and thus their nature) depends a little bit on how cool you are (I would not care about bias, but I am rather alone in this regard, I think) and your sample size. The sample size is a little dependent on what you can tell about the expected intra-subject variability. You need to be a little more detailed than just mentioning it is highly variable.

I usually do not recommend to use a pilot for evaluation if one of a number of Test treatments is a potential match for the Ref. This is because of the uncertainity. Some weirdo published a paper about this phenomenon, I think.

Pass or fail!
ElMaestro
lukamar
☆    

Poland,
2021-07-20 10:54
(982 d 13:25 ago)

@ ElMaestro
Posting: # 22480
Views: 1,996
 

 HVD Pilot: 3 candidates, 1 reference

Hi ElMaestro,

❝ I usually do not recommend to use a pilot for evaluation if one of a number of Test treatments is a potential match for the Ref. This is because of the uncertainity. Some weirdo published a paper about this phenomenon, I think.


Can you please elaborate about this uncertainity more? Where is it coming from? Can you reference this weirdo's paper if possibile?
ElMaestro
★★★

Denmark,
2021-07-20 11:31
(982 d 12:49 ago)

@ lukamar
Posting: # 22481
Views: 1,973
 

 HVD Pilot: 3 candidates, 1 reference

Hello Lukamar,

❝ Can you please elaborate about this uncertainity more? Where is it coming from? Can you reference this weirdo's paper if possibile?


Let us say you do a pilot trial with two test products and one Ref. in N=12 subjects.
For T1 versus R you get a 90% CI of 91.76%-118.23% for a metric of interest, like Cmax or AUCt.
For T2 versus R you get a 90% CI of 88.33%-114.99%.

One might be inclined to think that T2 is the better match. The numbers above would suggest it since PE is closer to one for T2 than for T1, right? So perhaps you select T2 as the candidate formulation for the pivotal trial. But it turns out that this selection is associated with a lot of uncertainty.
  1. T1 may actually be a lot better than T2; even if the figures above suggest the opposite. The probability of this being the case can't be worked out from the figures (but trust me, it is regularly unpleasantly high)
  2. The figures above suggest that T1 versus R has a GMR of 104.2% while T2 versus R has a GMR of 100.8%.
Perhaps you calculate a sample size for the pivotal trial assuming that GMR is not more than 5% off (like GMR=0.95), or even GMR=0.9 which will give you a higher sample size and more comfort. But this assumption is with a high probability wrong (and again, you can't derive the chance that you are wrong from the figures above).

It all comes down to the true performance of T1 and T2 versus R. This is the one thing you will not know, ever. I used the terms "high" and "a lot" above; if you tell me the true performance CV and GMR of T1 versus R, and T2 versus R, then I will be happy to quantitatively define "high" and "a lot".

In essence, such issues are described here and here.
These papers will give you an idea of making the wrong decision if you try to select one out of two.

Pass or fail!
ElMaestro
UA Flag
Activity
 Admin contact
22,957 posts in 4,819 threads, 1,636 registered users;
81 visitors (0 registered, 81 guests [including 7 identified bots]).
Forum time: 23:20 CET (Europe/Vienna)

Nothing shows a lack of mathematical education more
than an overly precise calculation.    Carl Friedrich Gauß

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5