Sample Size Estimate in BE Studies [Power / Sample Size]

posted by ElMaestro  – Denmark, 2022-05-13 04:02 (259 d 08:34 ago) – Posting: # 22984
Views: 760

Hi Sereng,


(i) The sample size for a parallel trial will generally be somewhat higher. This is because the between-subject variance is most often higher than the within-subject variance (anomalies exist, but they have to do with the estimation methods). When you do a parallel trial, the (residual) variance estimate incorporates both the withins and betweens, and both are positive quantities.

(ii) This is not always straightforward. If you have CVs from data on the log scale (yes, on the log scale) for the two treatments then you can convert them to useful variances. Variances can handily be pooled (weighted according to sample size in the two groups). Most often you just aim for balance between groups, so the average of the two variances serves well. Your resulting variance estimate can now be plugged into the software of your choice along with a guess for the match, like 95% similarity of Cmax or AUCt.
Your software may prefer CVs as argument and it may even accept a vector of variances or CVs as argument, so you don't need to pool.
To convert between CV and variance, use CV = sqrt(exp(Variance)-1)

But you really need to start out with CVs based on data on the log scale.

Pass or fail!
ElMaestro

Complete thread:

UA Flag
Activity
 Admin contact
22,477 posts in 4,708 threads, 1,603 registered users;
25 visitors (0 registered, 25 guests [including 9 identified bots]).
Forum time: 11:36 CET (Europe/Vienna)

In these matters the only certainty is
that nothing is certain.    Pliny the Elder

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