Bioequivalence and Bioavailability Forum

Hi M Murphy,
I have added some comments below. Best, -Dave

❝ Is (uT-uR) the geometric mean ratio (point estimate) that has to be within 80 - 125 % and also the value that has to be within the limits your program produces?!?

Yes, on the ln scale, uT-uR is mathematically the same as uT/uR, the point estimate.

❝ If I have a three period two sequence (TRR and RTR) partial crossover study design, is the within-subject reference variance calculated using all the reference data or just period three reference data?

All of the reference data.

❝ Is uR the mean of all of the reference data (from all periods and sequences) or just the mean of the references in the first two periods?

It is fit using SAS, and uses all the data. Please note that SAS does a model fit, and the number might be mathematically different than the straight average.

❝ Thus is the replicated reference solely for calculation of the within-subject reference varaince by which the confidence limits are widened (then period 3 data is forgotten about) and the confidence intervals calculated using a two period crossover study as normal?

No, data in all periods are used. There is an FDA guidance which describes it:
http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM209294.pdf

You really need a biostatistician to do this properly.

❝ Any help is much appreciated!

No problem!