4 replies to this topic

[Susana Carvajal]

Hi, I am looking for help regarding the execution ANOVA model in WinNonlin 6.4 of three-way cross-over bioequivalence study (3 period, 3 sequence, 3 treatment)

 

Used design

 

                P1          P2          P3

S1           T1           R             T2

S2           R             T2           T1

S3           T2           T1           R

 

I strongly appreciate if someone could help me about this matter.

 

Thank you for your attention.

 

Best regards

[Simon Davis]

Susana, my understanding is that it is possible to do a three-way crossover bioequivalence (BE) analysis in WinNonlin, provided that all sequences are represented, and the subjects are evenly divided into each possible sequence group.  However your dataset does not appear to meet these requirements.   Typically a Williams' design (3-treatment, 3-period, 6-sequence) is suggested, probably it’s best you go back to the statistician who wrote your SAP and discuss with them what they intended to analyse with the design.  This thread may be of use to you.

 

http://forum.bebac.a...escasc=DESC#p54                        

In the mean time I've forwarded this thread to a more statistically qualified colleague to see if they can add anything.  If you need some more statistical support you may be looking at consulting, in which case we will follow up via the support case you also opened on this issue and will require a copy of the project/dataset.

 

 Simon

[Helmut Schütz]

Hi Susana,

 

as Simon wrote some (including myself) prefer a six-sequence Williams’ design. However, if you don’t want to deal with first-order carry-over – a model which is considered obsolete by many – nothing is wrong with your Latin Square. Williams’ designs are mentioned in some regulatory guidelines (EU’s EMA, Brazil’s ANVISA, Australia, Russia,…).

 

The setup in Phoenix/WinNonlin is straightforward anyhow. In the BE object | Model | Reference Formulation select R. BE will be calculated for T1 vs R and T2 vs. R. No big deal.

[Susana Carvajal]

Hi Susana,

 

as Simon wrote some (including myself) prefer a six-sequence Williams’ design. However, if you don’t want to deal with first-order carry-over – a model which is considered obsolete by many – nothing is wrong with your Latin Square. Williams’ designs are mentioned in some regulatory guidelines (EU’s EMA, Brazil’s ANVISA, Australia, Russia,…).

 

The setup in Phoenix/WinNonlin is straightforward anyhow. In the BE object | Model | Reference Formulation select R. BE will be calculated for T1 vs R and T2 vs. R. No big deal.

 

Thank you very much for your answer,

SC

[Susana Carvajal]

Susana, my understanding is that it is possible to do a three-way crossover bioequivalence (BE) analysis in WinNonlin, provided that all sequences are represented, and the subjects are evenly divided into each possible sequence group.  However your dataset does not appear to meet these requirements.   Typically a Williams' design (3-treatment, 3-period, 6-sequence) is suggested, probably it’s best you go back to the statistician who wrote your SAP and discuss with them what they intended to analyse with the design.  This thread may be of use to you.

 

http://forum.bebac.a...escasc=DESC#p54                        

In the mean time I've forwarded this thread to a more statistically qualified colleague to see if they can add anything.  If you need some more statistical support you may be looking at consulting, in which case we will follow up via the support case you also opened on this issue and will require a copy of the project/dataset.

 

 Simon

 

Thank you very much for your answer,

 

SC