Bioequivalence and Bioavailability Forum

Dear All,

within the scope of bioequivalence studies one has usually not to deal with any 'center effects'. This is simply because BE studies are usually performed at one center only, monocentric.
But usually ... is not always ever .

Thus in the near future I have to deal with my first multicenter cross-over bioequivalence study, moreover with patients instead of healthy volunteers.
Since the guidelines do not deal with multi-center studies to a great extent (not to say the extent is not significant different from zero ) I have some uncertainties.

My questions:

• Is the inclusion of center effects really mandatory?
  Since the protocol and its implementation will be identical in the centers I do not expect any difference. Moreover the centers are from the same country so that the patient population is not expected to be substantial different also.
  
• Is anybody out there who has experiences of regulatory acceptance of evaluations without / with inclusion of center effects?
  Astonishing enough I have reviewed a lot of literature abstracts (Google) that mention only the usual evaluation without to mention the examination of any center effects.
  
• If center effects are mandatory, should we treat them as fixed or as random effects?
  The rationale behind that multicenter design is only to accelerate the number of patients accrual. We are not so interested to generalize to a "population" of centers. With the planed 2-3 centers this would also a little bit hazardous I think.
  
• How should the ANOVA model be formulated?
  Which interaction terms?
  Which nested factors (subjects nested within centers and nested within sequences ...)?
  Remove non-significant interactions in the final evaluation?
  (Best would be if I could receive ready to go SAS code for that, ASAP .)

Experiences and opinions on this topic are highly appreciated.
BTW: The study is aimed to the European regulators.