Bioequivalence and Bioavailability Forum

Dear All,

I'm looking at a set of slides recently presented by Dr. Schütz (thank you for your great work!) in Ljubljana. I have a doubt regarding the model to be used for the analysis at the end of Stage 2 when the two-stage approach proposed by Potvin et al. is adopted. In the slides, the following model is reported at page 60:

- Fixed: treatment + period + sequence + stage*treatment
- Random: subject(sequence*stage)

Here I will not consider the issue fixed vs. random for the subject effect, but I would like to focus on the following problem: which effects should be included in the model?

In the paper by Potvin et al., the estimates of sigma² to be used for calculating the confidence interval at the end of Stage 1 and Stage 2 are indicated at page 253 as s²₁ and s²₂, respectively. On the same page it is stated that "the error term derived from the GLM ANOVA model including sequence, stage, period(stage), treatment, subject(sequence*stage) will give the appropriate s² term". To me it seems that this is the model that should be used at the end of Stage 2. From the different model reported in the slides by Dr. Schütz, a quantity indicated as SSmean in the paper can be obtained as the sum of squares (SS) of the stage*treatment effect (page 253). But this is only a component of s²₂, as described in the formulas at the end of page 252.

I would like to know your opinion on this point. Thank you very much for your help.