Bioequivalence and Bioavailability Forum

Hi Loky do,

❝ Now another case for a non highly variable drug, […]

What you confirmed already:

• Subjects were not erroneously administered two doses of T.
  
• Subjects are not poor metabolizers or “super-absorbers” (since you observed “normal” concentrations after R).
  
• No problems in bioanalytics.

What is still possible:

• A Subject-by-Formulation interaction, i.e., due to different properties of the products (excipients, manufacturing, …) T performs ‘better’ than R in these subjects – possibly belonging to a sub-population.
  However, an S×F-interaction is extremely rare (the ‘Two Lászlós’ doubt that it exists at all and is not just a statistical artifact). Without a replicate design cards are stacked against you (see the FDA’s guidance, Section C.2).

❝ Is it applicable to consider these subjects as outliers, is there any way to conclude this?

That’s similar to the cases we discussed before. Even if you would perform a statistical test (which is not acceptable acc. to current regulatory practice), you would get a p-value of 0.2453 which would speak against excluding them. Even for Health Canada (where an outlier assessment of studentized model residuals can be done if pre-specified in the protocol) you could exclude only one subject (≤5% if the sample size is ≥20).

❝ or could we re-dose them again to be confirmed?

Re-dosing is discouraged and likely not accepted.

❝ or simply the product fails ?

I’m afraid, yes.