Bioequivalence and Bioavailability Forum

Hi K2k,

❝ How its not accounted for period effects?

Again: Vinay’s infamous RRT1T2 is not randomized!
Imagine you have period effects of ±0, +20, +15, –10% and ‘true’ bioequivalence of both formulations, i.e., T₁=T₂=R=1. Responses in the four periods are expected to be P₁=1+0=1, P₂=1+1×20%=1.20, P₃=1+1×15%=1.15, and P₄=1–1×10%=0.90. Estimated treatment effects will be R=(P₁+P₂)/2=(1+1.20)/2=1.10, T₁=P₃=1.15, and T₂=P₄=0.90. GMRs will be T₁/R=1.15/1.10=104.55% and T₂/R=0.90/1.10=81.82%. That’s not the 100% we expect (biased due to lacking randomization).
Any cross-over design which is not randomized & balanced for treatments within periods cannot give unbiased treatment effects.

❝ As per the Link http://forum.bebac.at/mix_entry.php?id=10358 Mr. d_labes suggested sequence for the same but sequence as changed little bit.

If you have read the entire thread (and the linked references) you would have noticed that it is not that easy.

❝ Shall i consider this sequence or not.

Never apply anything you don’t completely comprehend.

---

PS: To quote Detlew:

“I myself would not go with such a design due to the uncertainties involved.”

---

PPS: What if both tests have similar GMRs? With which one would you perform the pivotal study? Would be nice to get some information about the products’ variances. Maybe we can work in the spirit of Balaam’s design?

T1 R  R  T2
R  T1 T2 R 
R  T2 R  T1
T2 R  T1 R 
T2 T1 T2 T1
T1 T2 T1 T2

Don’t ask me how to evaluate that.
Though I have some ideas…
_{Homework/exercise: Is the design balanced for period effects? And for carry-over?}

Why don’t you simply run two separate fully replicated three period studies – and tell us afterwards how much CV_WR differed between the two?

T1 R  T1     T2 R  T2
R  T1 R      R  T2 R