Replicate Design ≠ Higher-Order Cross­over [Design Issues]

posted by Helmut Homepage – Vienna, Austria, 2023-03-14 11:37 (458 d 04:15 ago) – Posting: # 23499
Views: 3,199

Hi Sereng,

❝ … my associate is using the term "full replicate" to describe a four-way crossover BE study of Test (Fed), Test (Fasted), Reference (Fed) and Reference (Fasted), with each treatment given only once. In my view, this is not an appropriate us of the term "full replicate" but I want to be certain.

You are right. For some of the designs see the vignette of the R package replicateBE.
For the named food-interaction study you have three options. Primarily you are interested to demonstrate BE in fasted and fed state. The assessment of food effects of test and reference is just ‘nice to know’.
  1. Two separate 2×2×2 studies AB|BA and CD|DC, where A = Tfed, B = Rfed, C = Tfasting, D = Rfasting. Esp. for controlled release products BE in fed state is more difficult due to potentially different food effects of T and R. Hence, I suggest to perform this study first.
  2. A higher-order crossover in a Williams’s design. You can use any of the six designs ADBC|BACD|CBDA|DCAB, ADCB|BCDA|CABD|DBAC, ACDB|BDCA|CBAD|DABC, ACBD|BADC|CDAB|DBCA, ABDC|BCAD|CDBA|DACB, or ABCD|BDAC|CADB|DCBA. What you must not use is the Latin Square ABCD|BCDA|CDAB|DABC. Only from a Williams’ design you can extract balanced pairwise comparisons (see there).
  3. Like #1 but in the same subjects and the second following immediately after the first. Essentially it’s one study with two parts.
#1 is most simple and allows to perform both studies in either fixed-sample or adaptive two stage designs. However, assessment of food effects is done in parallel groups and therefore, has low power.
Whilst #2 is popular, there are drawbacks. If variabilities in fasted and fed state are different (commonly fed > fasted), you have to base the sample size on the worst case.
#3 is for the courageous ones (though all of my studies were accepted by European agencies :-D). Both parts (fed and fasting) are TSDs like in #1. You start with a best guess sample size for the condition with lower variability. Food effects can by assessed as paired comparisons (assuming identical period effects), which is much more powerful than the parallel comparison in #1.

❝ Additionally, what is the appropriate term, "Full Replicate" or "Fully Replicate"?

You can either say “full replicate design” or “fully replicated design”. In the former term you use an adjective and in the latter adverbs.

Dif-tor heh smusma 🖖🏼 Довге життя Україна! [image]
Helmut Schütz

The quality of responses received is directly proportional to the quality of the question asked. 🚮
Science Quotes

Complete thread:

UA Flag
 Admin contact
23,056 posts in 4,840 threads, 1,641 registered users;
147 visitors (0 registered, 147 guests [including 65 identified bots]).
Forum time: 16:52 CEST (Europe/Vienna)

Exploratory analysis: The art of finding a Rembrandt
in a Jackson Pollock.    Stephen Senn

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz