Bioequivalence and Bioavailability Forum

Dear all,

after almost two years of rumors, today a 'Question and Answers Document' to the European BA/BE-Note for Guidance was published by EMEA - dated 27 July 2006.

I added the link to the Guidance page; here is the direct link (118kB PDF) to the document as well.

Main points of the document are

1. Introduction
   
2. Wider acceptance range for C_max (generally discouraged, only in rare cases, limited to 0.75–1.33)
   
3. Outliers. Forget it.
   
4. One confidence limit outside 0.80 or 1.25: not bioequivalent, but must be seen within the pharmacokinetic / pharmacodynamic context.
   
5. Nonparametrics for parameters other than t_max. See my personal remarks below
   
6. Use of metabolite data (still ambiguous conditions ar given).
   
7. Widening of C_max-limits for metabolites (same conditions as given in #2 but even more restrictive).
   
8. Definition of a HVD / HVDP (30% intra-subject CV).
   
9. Selection of strenghts for drugs with non-linear PK.
   
10. Urinary PK data in BE.
    
11. Standardization of food intake.

My personal favorite is Section 2 concerning C_max:

‘The possibility offered here by the guideline to widen the acceptance range of 0.80 – 1.25 for the ratio of Cmax (not for AUC) should be considered exceptional and limited to a small widening (0.75 – 1.33).’

… which is in contradiction to the NfG, which states for the AUC-ratio:

‘The 90% confidence interval for this measure of relative bioavailability should lie within an acceptance interval of 0.80–1.25. […] In rare cases a wider acceptance range may be acceptable if it is based on sound clinical justification.’

So by a little sidestep widening of the acceptance range for AUC is now prohibited!

Another nice one comes from Section 5:

‘Parametric testing using ANOVA on log-transformed data should be the rule. Results from non-parametric statistical methods or other statistical approaches are nevertheless welcome as sensitivity analyses. Such analyses can provide reassurance that conclusions from the experiment are robust against violations of the assumptions underlying the analysis strategy.’

The problem with ANOVA is the non-robustness against violations of assumptions. IMHO it’s simply a bizarre idea to show robustness of ANOVA with a nonparametric analysis.