Bioequivalence and Bioavailability Forum

Hi BE-proff,

❝ Let's say I have the followind results after typical BE-study:

❝ n=30, CV=0.20, ratio=0.97, UL=1.02 and LL=0.92

The study passed by a nice margin. Why are you interested in the irrelevant post hoc power? BTW, I cannot reproduce your 90% CI:
library(PowerTOST)
round(CI.BE(alpha=0.05, pe=0.97, CV=0.2, n=30), 4)
 lower  upper
0.8892 1.0582
Only with a strange α:
round(CI.BE(alpha=0.147, pe=0.97, CV=0.2, n=30), 2)
lower upper
 0.92  1.02

1.

❝ if I want to calculate power using default settings of PowerTOST I see:

❝ power.TOST(CV=0.20, n=30)

❝ [1] 0.948603

Sure. Type help(power.TOST) in the console. You will find:

theta0

Defaults to 0.95.

theta1

Lower bioequivalence limit. Defaults to 0.8.

theta2

Upper bioequivalence limit. If not given theta2 will be calculated as 1/theta1.

So you are essentially asking which power can be expected with a CV of 0.20, assuming theta0 0.95, and the conventional acceptance range of 0.8–1.25. Works as designed.

2.

❝ If to add ratio I have:

❝ power.TOST(CV=0.20, n=30, theta0=0.97)

❝ [1] 0.9781158

That’s the post hoc (aka a posteriori) power of the study.

3.

❝ If to use also limits of CI result the result is:

❝ power.TOST(CV=0.20, n=30, theta0=0.97, theta1=0.92, theta2=1.02)

❝ [1] 2.334844e-05

Now you asking for the power to show BE where the acceptance range practically equals the CI in the study. Of course this chance is very low.

❝ Do I use formulas correcly and which option is better?

1. Maybe. But it gives you only the power for another study with the same CV and sample size and theta0 0.95. Maybe more interesting: pa.ABE(CV=0.2, theta0=0.95, targetpower=0.9)
   
2. Meaningless.
   
3. Wrong.

---

Congratulations for your post #100!

Hi BE-proff,

to say it in a different manner, but stating the same:
Q: How to calculate power correctly?
A: By doing it before the trial (especially if 'correctly' means 'meaningfully').