Bioequivalence and Bioavailability Forum

❝ whether can we perform statistical analysis for 'x' in the first 22 subjects from 1 to 22 and for 'y' in all 44 from 1 to 44?

Dear Sasikumar,

I think it could be OK to analyse x on basis of the first 22 completers, but I would advise against it, because:
1. You have better power with 44 than with 22 subjects. If you end up with a CI from 79% to 113% you would wish you had taken 26 for analysis and not 22, etc.
1½: It is unclear why you would go for lower power when the subjects are in trial anyway, unless of course we talk Danish requirements for BE where you are spanked, keelhauled, and ridiculed if you have too high power.
2. You could run into some issues regarding the way you are handling the 22 hot subjects. Let's say for example that you are ethical and draw a smaller blood volume from 22 subjects because you don't need to analyse x on those. This would give rise to potential grouping issues (i.e. an extra factor in the model+anova for y), discussed just recently in this forum, and you should randomise the 22 from the 44. Let's in stead say you draw the same amount of blood in all 44, but only analyse 22 of them for x. In that case you need to make sure the 22 chosen ones are a random sample of the 44, so this should, as far as I can grasp, be done at the time of seq randomisation and deffo described well in the protocol. I think I would even go as far as to randomise 11 from seq TR into the pool that are analysed for x and 11 from seq RT, just to try to aim for sequence balance.

EM.