Bioequivalence and Bioavailability Forum

Mikalai
★

Belarus,
2018-10-31 15:04
(1857 d 05:09 ago)

Posting: # 19510
Views: 2,456

Full-replicate design in two groups [General Statistics]

Dear all,
We are conducting a full replicate design study. Due to recruiting difficulties we had to split the study in two groups. It has been suggested that we should change our statistical model to model II of FDA. What risks, if any, carry on this model to our bioequivalence? Can additional factors in model artificially reduce our CV and push us out of the scaled approach to usual 125%-80% one (I am not a statistician, so my question may be statistically correct)? Should we change our usual 4-factor model?

mittyri
★★

Russia,
2018-11-01 13:08
(1856 d 07:04 ago)

@ Mikalai
Posting: # 19514
Views: 2,073

Full-replicate design in two groups

Post reply

Dear Mikalai,

❝ It has been suggested that we should change our statistical model to model II of FDA.

Sorry, but who is the author of suggestion?

❝ What risks, if any, carry on this model to our bioequivalence?

This approach looks strange to me. If you are basing on FDA models, why are you using FDA Model II and ignoring other criteria given in Progesterone Guidance?
See also similar question and Helmut's answer here

❝ Can additional factors in model artificially reduce our CV and push us out of the scaled approach to usual 125%-80% one (I am not a statistician, so my question may be statistically correct)?

Resulted CVintra can go up or down. There are many factors which can inflate on CVintra (widening it or squeezing)
Usually you won't see drastically changed CV, but I would not bet on that

❝ Should we change our usual 4-factor model?

Whenever possible: keep EMA ABEL clean!

—
Kind regards,
Mittyri

Full-replicate design in two groups [General Sta­tis­tics]

Full-replicate design in two groups

Full-replicate design in two groups [General Statistics]