Bioequivalence and Bioavailability Forum

Dear joyjac!

Whereas the standard 2×2×2 cross-over design (two treatments, two sequences, two periods) gives us the inter-subject- (between subjects) and intra-subject- (within subjects) variability, replicate designs (or sometimes called 'higher-order' or 'optimal' cross-over designs ) give us the additional information on the variability of treatments.

The most simple form is the two-sequence dual (two sequence, two treatment, three period) design [1] (sometimes called 'extra-reference' design):
S1: T R R
S2: R T R

A more sophisticiated design is the two-sequence, four-period design [2]:
S1: T R R T
S2: R T T R

Designs [1] and [2] are given in FDA's guidance; others (e.g. Balaam's design and the four-sequence, four-period design) are discussed in the literature, but to my knowledge rarely used.

Pros:

• desirable statistical properties (i.e. allow for the estimation of treatment variability)
  
• reference-scaled average BE may protects against a unreasonable high sample size in 2×2×2 based on a 'bad' reference product; recommended for highly variable drugs (CV>30%) in various guidelines (e.g. ANVISA)
  
• it may give protection against outliers based on product failure (distinguishes from subject-by-formulation interaction)

Cons:

• errors in the randomization procedure are more likely
  
• additional periods are time-consuming
  
• there is a higher chance of drop-outs
  
• the number of blood-samples per subject may be rather high

So if you are dealing with an HVD, it may be a better option to go for a reference-scaled ABE in a conventional 2×2×2 design (Tothfalusi, L. and L. Endrenyi; Limits for the scaled average bioequivalence of highly variable drugs and drug products. Pharm Res 20(3), 382-389, 2003). You will find some slides at FDA's website. If you want to be protected against outliers caused by product failure of the reference, you may go for a replicate design (prefereably [1]).

regards
Hermann Rotter

---

Edit: Link corrected for FDA’s new site. [Helmut]