Bioequivalence and Bioavailability Forum

Dear Matz!

❝ thank u so much for your reply...

Welcome!

❝ One more clarrification please....U said Proc Mixed is for >2treatments but FDA guideline "Statistical Approaches to Establishing Bioequivalence" says PROC MIXED in SAS should be used for the analysis of replicated crossover studies for average BE[p.13].

❝ Is there any particular reason behind this?

Example of a [TRTR – RTRT] replicate design:
If a subject in the second sequence drops out in the last period you have data from three periods; two values of the reference (from periods I & III), and one value of the test (from period II).
Proc GLM drops the entire subject, but Proc Mixed uses all available data.

❝ Because 2 treatment 4 period design is a replicate crossover design. I compare a sample data set of 2x4 design with missing values (example 4.4 in the text "BIOEQUIVALENCE and STATISTICS in CLINICAL PHARMACOLOGY") using both GLM and MIXED procedure. both will give same output. then why FDA says to use MIXED procedure for replicate crossover design?.

❝ Any reason?.....

OK, you are right - my answer was not correct (I had some type of Williams' design in mind).
It should read 'Proc Mixed for >2 periods'.

❝ someone says GLM procedure deletes the entire subject from an analysis if that person has any missing data, but MIXED does not. Is it correct?

Yes, this is correct (see above).
Sorry for my confusing previous answer.

Regards,
Hermann