Bioequivalence and Bioavailability Forum

Hi Lucas,

❝ ❝ […] such a value might be physiologically impossible – especially if the value is embedded by two high concentrations.

❝

❝ […] afaik there is the possibiility of an enterohepatic circulation, for example, which would make possible for the drug concentrations in plasma to have such "triangle" behaviour.

Correct, but enterohepatic recirculation leads to secondary peaks – not drops. Since we are “talking BE” such a behavior is known beforehand and f.i. sampling should be adjusted accordingly. Isolated peaks are then more unlikely.

❝ Also, statisticians are mostly not in favor of outlier exclusion, and such points are outliers themselves. It's more important to understand why it happened than to just exclude it.

Correct again. It is a pity that some statisticians have limited knowledge of PK.
Statistics is just a tool. Many issues could be avoided if pharmacokineticists, bioanalysts, and statisticians talk more to each other already in designing studies.

❝ It could have happened due to an accidental swap of samples, […] high LLOQ or other unknown reason...

IMHO your first case is the most common one.
Remember that the LLOQ is obtained from spiked samples. It might be that in a particular sample (coeluting compounds leading to a different matrix effect) the actual LLOQ might be higher.

❝ […] we have more complicated situations like for endogenous drugs that after baseline correction might present lots of triagulations points.

Yes, but that’s a different story. If the measured concentration equals the baseline, you get after subtraction a “true” zero. Such values are valid and should be used. If C <LLOQ then force it to zero.

❝ ❝ I don’t like the common practice of repeating only the doubtful value. We always re-analyze the two neighbouring samples.

❝

❝ That would be the best way, but ANVISA does not allow reanalysis due to PK reasons, in other to avoid a manipulation of the results.

OK.