Bioequivalence and Bioavailability Forum

Dear Mathews,

❝ If my study needs total 26 subjects, how many subjects I want to take as

❝ additional subjects? 26+2 or 26+4?

❝

❝ Is there any special rule for calculating the number of aditional

❝ subjects?

No special rule that I know of, you should decide on a case-by-case basis:
- how long is your trial, and particularly the wash-out period ? The longer it is, the higher the risk that some subjects may not turn up in Period 2.
- what are the known side effects of your drug ? The more frequent and severe they are, the higher the risk of having to withdraw subjects. And particularly, does it cause vomiting ?

❝ Which type of randomization is best for a 2 x 2 crossover design?

❝ (I think block randomization is good)

❝

❝ If I use block randomization, what is the best block size? 2 or 4? why?

IMHO this is less important for a BE trial than for a double-blind, multi centre Phase III trial. In a Phase III you need to use block randomisation because you don't know how many patients each investigator will enroll and you want to limit the risk of unbalance. And you need blocks of at least 4 (if comparing two products) in order to limit the risk of breaking the blind (e.g. with blocks of 2, if you need to break the code for one patient, or if a patient experiences an AE which would be typical of one of your treatments, you would automatically also get the treatment of the other patient in the block). For a BE trial you already know how many subjects you will enroll and the trial is open-labelled. However most CROs go for block randomisation, with blocks of 4.

❝ In the randomization procedure, Is it necessery to generate separate

❝ randomization codes for total subjects (26) and seperate codes for

❝ additional subjects (2 or 4)? (by using different seed number in SAS

❝ pgm).

❝

❝ or is it possible to generate one randomization code by using one seed

❝ number for (26+2) or (26+4) subjects?

❝

❝ When we use this addtional subjects in the analysis?

Answering the last question will give you answers to the first...
I've seen different options used:
- use the data from all subjects who completed the trial (in your case, up to 28, if 26 + 2). Then I can't see a need for a separate code for additional subjects.
- use the first 26 subjects who completed the trial, and if one of the 26 did not complete for any reason, then use the stand-by subject with the same treatment sequence in order to keep a balanced design. Then you would need a separate code for additional subjects in order to be sure to have both sequences in your stand-by subjects.

But how do you intend to randomise 26 subjects with blocks of 4 ?

❝ In the case of block randomization, is there any difference in the result

❝ of analysis (90 % C.I and ANOVA table) if I use the following two methods[...]

I would say no, but I leave this one to HS !

Regards
Ohlbe