Bioequivalence and Bioavailability Forum

Dear S.Sasikumar!

To quote the first paragraph of the forum's policy:

We expect a basic knowledge on BE/BA or related fields and the willingness to begin first with the Search function for similar problems.

In BE we are not interested in rejecting the null-hypothesis of equivalence (this is significance testing), but to reject the null-hypothesis of inequivalence. This can be done by performing two one sided t-tests at alpha 0.05 (the first one 'looking' whether T is below 80% of R and the second one 'looking' whether T is above 125% of R) - the sum of both p-values must be <0.10 to claim BE. Inclusion of a 90% confidence interval within the acceptance range is actually the same thing.
In any study of a high enough sample size you will get a statistically significant difference, but this has nothing to do with clinical significance (which is set mainly to 20%, or to 10% for NTI drugs in some regulations).
If your study was properly planned in terms of power and you get a significant treatment effect there are a couple of explanations possible:

• lower intra-subject CV,
  
• less difference of test from reference,
  
• fewer drop-outs than expected, and
  
• any combination of the above.

This is just a short blow; I leave the details to the statistical experts.

Edit:
See this post, and this post for references to textbooks.

dear sasikumar!

this guideline may help in understanding the different concepts of testing for equivalence/non-inferiority and testing for difference/superiority: http://www.tga.gov.au/docs/pdf/euguide/ewp/048299en.pdf

testing for equivalence with a type I error of 5% can be done by the two-sided 90% confidence interval inclusion approach (similar to the two one-sided tests approach TOST). the calculated two-sided 90% confidence interval for the difference between test and reference treatment (e.g. ratio of geometric means) must fall completely within pre-specified margins of equivalence. conventional margins in pharmacokinetics are 0.8 to 1.25 which are symmetric margins for ratios as 1/0.8=1.25 and 1/1.25=0.8.

for more detail regarding testing for equivalence you may find this book helpful: Wellek S (2003). Testing Statistical Hypotheses of Equivalence. Chapman and Hall/CRC Press.

best regards

martin