Bioequivalence and Bioavailability Forum

Hi IP,

❝ Kindly let me know from where can I get the below mentioned parameters. I use Winnonlin for my BE calculations.

CV_intra is already given in WinNonlin’s standard output for 2×2 cross-over studies. For higher-order cross-over calculate it from the Var(residual) as

CVintra=sqrt(exp(Var(residual))–1).

MS_intra and SS_intra are not given in PHX/WNL’s standard output (I guess we will see an update in the next revision). Note that MSintra=log(CVintra²+1). You also need the degrees of freedom. Get them from the Partial Tests, row sequence*subject. The inverse χ² is not available in PHX/WNL’s custom transformations (a request is scheduled for the next revision of PHX).

❝ Can I do the CI calculations for ISCV in Excel?

Yes, but I would not recommend it (M$ Excel’s CINV() is known to be buggy; see also this post). I would suggest the free software and the famous package PowerTOST.
Example: 2×2 cross-over, 20 subjects balanced (n₁=n₂=10; df=n₁+n₂-2=18), CV_intra 20%.

library(PowerTOST)
100*CVCL(CV=20/100, df=18, alpha=0.05)

Which gives by default the upper one-sided interval
lower CL upper CL
 0.00000 27.94244
The example in the previous posts

100*CVCL(CV=15/100, df=14, side="2-sided", alpha=0.05)

gives
lower CL upper CL
10.95351 23.85428 ( post #126)
or better (since I would not use a two-sided interval any more)

100*CVCL(CV=15/100, df=14, alpha=0.05)

lower CL upper CL
 0.00000 22.03429 ( post #127)