Bioequivalence and Bioavailability Forum

Dear Imran!

❝ […] but sequence effects was not discussed up to the level.

Really not?

❝ Here I mean to say statistical significant Sequence effect. I was trying to rule out the statistical possiblities for the sequence effects.

I thought that I already gave you detailed answers above… Which are the open questions?

Dear Łukasz!

❝ ❝ OK, to be more precise, if:

❝ ❝ • the study was performed in healthy subjects,

❝ ❝ • the drug is not an endogenous entity, and

❝ ❝ • an adequate washout (no predose concentrations in period II) were maintained.

❝

❝ Is meeting these criteria enough to ignore sequence effect?

Actually this a quote from the FDA’s Guidance and (hopefully) consensual – Freeman’s paper disproving Grizzle’s 2-stage approach was published 19 (!) years ago.

❝ From reference given previously I can see that also equivalence must be present.

This is a doubtful reference. Anyhow, yes, if you want to demonstrate bioequivalence your CI must be entirely within the acceptance range. Statistical significant sequence effect (SE) or not.

❝ And in study that I'm currently participating in, it turned out that test drug did not meet BQ criteria for Cmax when compared to reference drug and significant sequence effect for Cmax was detected.

So the test formulation is not BE to the reference in respect to C_max.

❝ Now this publication [E Zintzaras, 2000] claims that:

❝ "The sequence effect is confounding with the unequal residual effect and with the formulation by period interaction. […]"

True.

❝ "[…] Since the existence of the sequence effect questions the quality of the trial, the applicant should provide possible explanations …"

Yes, but since a statistical significant SE may have different sources – which can’t be separated in a nonreplicated study – it’s futile to aim at an “explanation”. It’s only possible to speculate, which can’t be the basis of further statistical analyses!

❝ "… and information on the subjects, the trial conditions, the clinical settings and the assay methodology. […]"

True. Only by assessing the side conditions of the study (wash-out, randomization, lack of pre-dose concentrations, …) one can get some insights on the SE. The nasty thing is, that once the study is finished, it’s not possible to get any clues from the data. We can avoid a potential SE only in study planning. If a study is properly planned, any testing for a SE is rubbish.

❝ "[…] An additional statistical analysis on the data from the first period of the trial may support the bioequivalence. […]"

Nonsense. The paper was published in 2000 in a journal – though peer-reviewed – has an impact factor close to zero (about one year ago: 0.3). Obviously both author’s and reviewer’s “state of the art” was with Grizzle’s two stage approach of 1965 [sic]. Since then Freeman’s paper, meta-analyses, a book by Stephen Senn (which in between the lines deals to a good part with rejection of SE testing), chapters on the topic in books by Jones/Kenward, and Patterson/Jones were published. Although the author is with a European regulatory authority (which is known for its bizarre interpretation of the guidelines, like a posteriori power ≥80% needed for BE) I would not take the statement seriously, because:

• Patient’s risk in Grizzle’s method may be inflated
  
• If you are able to demonstrate BE in period 1 as a parallel design, your study was extremely overpowered; so your design is questionable…

❝ "[…] If it is proven that the sequence effect is a true effect then the generic may be approved for marketing authorization."

Only by means of “black magick” it’s possible to ‘prove the sequence effect as a true effect’ in a 2×2 cross-over study.

❝ So should we go deeper and try to find explanation for sequence effect? Is it possible at all to find anything if we pass criteria mentioned by HS (cited above)?

You can look at the design and the performance of the study, but if you are not BE, there’s no statistical “bail-out procedure” to deal with the problem. Sorry.