Bioequivalence and Bioavailability Forum

Hi Balaji,

❝ so we can use total variability instead of (inter subject variability and intra subject variability) for parallel design…

Not instead of… In a parallel design the total variability is all we have.
Again: Inter- and intra-subject variabilities are not accessible, only their sum!

❝ …using this 100√ℯVar(Residual)–1

I strongly advise to use the modified setup (see the “Phoenix WinNonlin 6.4 User’s Guide”, p85–6).

❝ …but winnonlin wont calculate total cv we have to calculate manually

Manually? Set up a custom transformation in the Data Wizard.

• Default model (not recommended)
  Final Variance Parameters [right click] Send To ► Data ► Data Wizard | Add Transformation
  Transformation Type: Custom
  Transformation: Custom Function
  New Column Name: CVp
  Formula: 100*sqrt(exp(Estimate)-1))
     Gives for my example above 80.54%.
  
• Modified model (adjusting for unequal variances and group sizes)
  Same procedure as above.
     Gives 46.23% for Reference and 111.34% for Test.