Bioequivalence and Bioavailability Forum

Hi SK,

❝ On review of p'cokinetic data it was observed that one of subject is showing highest T/R ratio for Cmax and AUC. So we just recalculate 90% CI by excluding that subject and those results were within acceptance limit. No specific outlier test is applied here.

If you have just pulled out the subject with the highest T/R without any further qualification then that won't be acceptable. If you have confirmed technical issues with the samples then you may be able to take out the subject, same thing if the subject vomited etc.

Dear Swapnil!

❝ On review of p'cokinetic data it was observed that one of subject is showing highest T/R ratio for Cmax and AUC. So we just recalculate 90% CI by excluding that subject and those results were within acceptance limit. No specific outlier test is applied here.

Caution: This might give assessors the impression of cherry-picking. Never fiddle around with evaluations – ‘just to see what happens’.
Have you ever heard the phrase “Don’t worry, it’s too late!”…?
You might consider sticking to the full data set and present the reduced data set as a sensitivity analysis - supported by the points below.

❝ Infact this is about fed biequivalence study for gastro-resistant formulation and importanly we have positive results for fasting study for the same formulation.

You are not alone. Quite often gastric-resistant formulations ‘behave’ nicely in fasted state and awful in fed state, not only caused by the formulation itself, but also due to GI-physiology. See the Q&A-document, 4. Bioequivalence of gastro-resistant preparations (e.g. omeprazole).

❝ What should be our future strategy in the same regard.

Follow the GL (the second point in my previous post) and the Q&A-document.

[…] only under the conditions that sampling times are designed to identify very delayed absorption and that the incidence of this outlier behaviour is observed with a comparable frequency in both, test and reference products, these incomplete profiles can be excluded from statistical analysis provided that it has been considered in the study protocol.

Only few studies are designed in such a way that they are able to ‘catch’ very delayed and/or low profiles to the same quality as earlier/high ones. For an example see this presentation (slide 41). Another point is ‘the incidence of this outlier behaviour is observed with a comparable frequency in both, test and reference products’. That’s a rather vague statement - what’s a ‘comparable frequency’? See the same presentation, slide 40. In a study in 24 subjects up to 4 outliers in one formulation (and none in the other) would be of comparable frequency (Fisher’s exact test, p>0.05).
You should consider a full replicate design in the future - you are in a much better position in an argument if you see the outlier behaviour only in one of the repeated administrations. A subject-by-formulation interaction could be ruled out in such a case.