Bioequivalence and Bioavailability Forum • Post hoc power & other things…

Post hoc power & other things… [Design Issues]

posted by Helmut – Vienna, Austria, 2013-02-20 21:11 (4868 d 12:50 ago) – Posting: # 10087
Views: 7,905

Dear Ken!

❝ Do you mean even we can not achieve the power (>80%) as long as BE is proven ??? Am I correct?

Post hoc power is completely irrelevant in bioequivalence. Either you demonstrated BE or not. Full :stop:

. Unfortunately the Malaysian authority asks for post hoc power sometimes (or always?). It’s time for them to learn. ;-)

❝ Nevertheless, the 2013 ASEAN Guideline on BA/BE will be out soon.

THX! Would you be so kind to post a link once the GL is published?

❝ ❝ If one expects a deviation of ±20% (c) the samples size has to be infinity even for a CV of zero. ;-) Should read “ie ±5%”.

❝

❝ I am sorry. Could you please kindly elaborate the last sentence. "If one expects a deviation of ±20%, the samples size has to be infinity even for a CV of zero. Should read “ie ±5%".

The wording of the Q&A document is nonsense, whereas the 2004 GL is correct:

The number of subjects required is determined by

the error variance associated with the primary characteristic to be studied as estimated from a pilot experiment, from previous studies or from published data,
the significance level desired,
the expected deviation from the reference product compatible with bioequivalence (delta, ie percentage difference from 100 %) and
the required power.

“Delta” mentioned in (c) is the expected deviation of test from reference. In many cases 5% (a T/R-ratio of 95%) is used. If you expect a deviation of 20% (ratio 80%) it is simply very, very unlikely to show BE – even if you include the entire population of our planet in the study… OK, the actual ratio in a given study might be >80% just by chance, but you cannot seriously plan a study for such a deviation. Examples (very, very low CV of just 5%):

library(PowerTOST)

sampleN.TOST(CV=0.05, theta0=0.8)

+++++++++++ Equivalence test - TOST +++++++++++

            Sample size estimation

-----------------------------------------------

Study design:  2x2 crossover 

Error: Null ratio 0.8 not between margins 0.8 / 1.25!



sampleN.TOST(CV=0.05, theta0=0.8000021)

+++++++++++ Equivalence test - TOST +++++++++++

            Sample size estimation

-----------------------------------------------

Study design:  2x2 crossover

log-transformed data (multiplicative model)



alpha = 0.05, target power = 0.8

BE margins        = 0.8 ... 1.25

Null (true) ratio = 0.8000021,  CV = 0.05



Sample size (total)

 n     power

4480622660   0.800000

↑ Would require more than the entire population of Asia (4.26 billions).

—
Dif-tor heh smusma 🖖🏼 Довге життя Україна!
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

Complete thread:

Trifluoperazine Ken Peh 2013-02-16 16:26
- Trifluoperazine drgunasakaran1 2013-02-17 08:53
  - Trifluoperazine Ken Peh 2013-02-18 07:28
    - Pilot or Two-Stage Helmut 2013-02-18 13:53
      - Pilot or Two-Stage Ken Peh 2013-02-18 18:19
        
        Pilot or Two-Stage Helmut 2013-02-19 13:37
        
        Pilot or Two-Stage Ken Peh 2013-02-20 03:34
        
        Post hoc power & other things…Helmut 2013-02-20 20:11
        
        Post hoc power & other things… Ken Peh 2013-02-22 08:17
        
        Post hoc power & other things… Helmut 2013-02-22 14:15
        
        Post hoc power & other things… Ken Peh 2013-02-24 09:08
        
        estimation ≠ calculation Helmut 2013-02-24 17:09