Prevent execution of LME with insufficient data

Phoenix Certara Forums Phoenix WNL basics
1

Dear all,

see the subject line…
I want to update the reference-scaling templates for the EMA’s methods. Recently the WHO started a pilot phase exploring reference-scaling for AUC. The WHO requires a 4-period full replicate design (TRTR|RTRT) in order to compare swT with swR. Easy to set up the fixed effects model Sequence+Subject(Sequence)+Period as given in the EMA’s Q&A document. As usual with the EMA, superfluous terms. The simple Subject+Period gives an identical residual variance. But we want to make regulators happy, right?

  • Include subjects with two administrations of R, calculate swR as sqrt(Residual Variance) and
    CVwR as sqrt(exp(Residual Variance)–1).
    Required for the WHO and widening of the BE-limits if CVwR >30% according to [L,U] = exp(∓0.76swR).
  • Include subjects with two administrations of T, calculate swT as sqrt(Residual Variance)
    Required for the WHO.

Works as a charm if the design is one of the currently implemented full replicates – and a couple of new ones as well. But the template should be able to handle all replicate designs (i.e., also the partial replicate TRR|RTR|RRT and the extra-reference TRR|RTR). For these designs I got the ERROR 11068 for #2. As expected, since T is not repeated and the remainder of the template is not executed. Any ideas to bypass #2?

2

Dear Helmut,

yes, you cannot do that directly,

I made some trick, hope it will make sense to you.

Works fine with my general template attached

The project is made in PHX7.

Hope it helps,

Mittyright

EMA ABEL Project Template v1_4_3_4_executed.phxproj (5.9 MB)

3

Hi mittyright,

you made my day!

I does. I also had the idea to introduce a dummy variable but didn’t know how.

THX a lot!

4

Dear Helmut,

I am working on a study where the original design was for two sequences, ABAB/BABA but due to dosing errors at the site, two different sequences were also applied to some subjects. They were AAAB and BBAA (A is the test and B the reference). So there are 4 sequences in total. Do you have a modifed version of the PHX RSABE template that can handle this design?

Thanks,

Dan Weiner