Bioequivalence and Bioavailability Forum

Dear Pavan!

❝ 1 T1 T2 T3

❝ 2 T2 T3 R

❝ 3 R T1 T3

❝ 4 T3 R T1

I guess copy-and-paste hit you - having T3 twice in period 3, and missing T2. I would go with:
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
            Period
          ───────────
Sequence  I   II  III
─────────────────────
   1      T1  T2  T3 
   2      T2  T3  R
   3      T3  R   T1
   4      R   T1  T2
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬

❝ Here some subjects are not receiving reference product.

True.

❝ Now my doubt is how to show the bioequivalence?

In any kind of BE model you compare essentially means and variances - not individual subject ratios. Degrees of freedom in the BIBD above are 2×(n-1) as opposed to 3×(n-3) in a 4×4 Williams' design (layout) - variance is 50% higher. In any BIBD design you pay a price for fewer periods - lower statistical power for an equal sample size.

I would not suggest a BIBD for four treatments. Your sample size must be much quite large! Four periods are commonly used in replicate designs - so I don't see ethical problems due to blood loss. Only with 5+ treatments I would go with a BIBD.