Bioequivalence and Bioavailability Forum

Dear Ratnakar!

❝ […] how can we should handle when there is a significant increase or decrease (aberrant or some times value below LOQ in between two values) in the concentration in the elimination phase (or any other phase like early absorption also) …

OK, that’s called an “irregular profile”. See the last example in this post. Due to overregulation many people abandon scientific judgment. Unlike for example in environmental analysis, in PK samples are not independent. For a single dose oral administration we know that the concentration at t=0 should be zero, increase until t_max, and decrease in some kind of exponential manner until LLOQ is reached. Since I’m a dinosaur of bioanalytics (manual injection of thousands of samples on packed-column GC and early HPLC equipment) I remember quite well, that every analyst looked routinely at the time-course of results before final release. Right now analysts - punished by badly educated inspectors – follow blindly foolish SOPs in a formalistic way which mainly allow for repeated analysis if something obvious happened (vial broken, spoiled extract, ‘bad’ chromatography). The most weird I experienced was a (very, very!) large CRO, where the analyst refused to look at the time-course of data, and the pharmacokineticist refused to perform a plausibility review before final release of analytical data.
I suggest to alter your SOPs in such a way that analytical data after review by the QAU are relased for review by an pharmacokineticist in a blinded manner. Then reanalysis may be initiated; this cycle is repeated until ‘strange’ results are clarified (original results confirmed, rejected, or no more sample left). Remember: ANVISA limits the number of reanalyzed samples to 20% of total.

❝ … if we notice that at the time of statistics, can we ask to reanalyze the particular time point of that subject. Also what can be considered as a base for such request?

A request for reanalysis after unblinding is a very bad idea and must be avoided if possible. The impression of ‘cherry picking’ may arise!

❝ […] how to deal with one or two time points in one or two subjects?

That’s a question not easy to answer… Science-driven CROs state procedures in the SAP (Statistical Analysis Plan), the clinical protocol, or an SOP. I would recommend the SAP or protocol, because it’s more transparent and allow to answer questions more quickly (SOPs generally are not included in the study report). It’s always possible that something went wrong with a particular sample, which will give you the erroneous result even after many repeated analyses. Most common reasons: contamination of the sample vial or stopper, forgotten stabiliser, coeluting compound leading to a matrix effect. In such a case it should be possible to eliminate the result from the profile. However, such a decision is not based on statistical methods but scientific knowledge and experience, great care should be taken rejecting and / or substituting questionable values.
Depending on the software you use, options to interpolate should be suitably set (e.g., linear interpolation in increasing part, and log/linear in the decreasing part of the profile).