Bioequivalence and Bioavailability Forum

Hi BEeman,

❝ in the new EMA-modified release guideline is stated, that […] The time point for truncationing should be based on the PK-Profile …

Bad science, IMHO. In my studies I followed FDA’s approach, i.e., based the cut-off time point(s) on clinical grounds (example: methylphenidate 3 h fasting, 4 fed; zolpidem: 1.5 h both). For some formulations it is very difficult to dis­tinguish “phases” in PK-profiles. Works in some subjects, but not in others. Mean profiles are completely useless. I have seen different studies with the same (!) reference where PK-profiles would lead to different cut-offs.
I know one example where the EMA in a scientific advice referred to one parti­cu­lar published study of the reference where (I would say by chance) sort of a “trough” was seen around two hours in the arithmetic mean curves. In many other [sic] studies the trough was seen at a substantially later time or not clear at all (only sort of a “shoulder” in absorption). The sponsor had to pay a high price for this bad advice. Due to the extreme variability in the early part the sponsor had to perform a replicate design with ABEL. With a later cut-off a simple crossover would have been sufficient…

Since the EMA (the final GL was adopted in Nov 2014!) didn’t publish the comments on the drafted guide­line so far we will never understand which reasoning hides behind their eter­nal wis­dom.

❝ … and should be specified in the study protocol.

Sure.

❝ Considering following biphasic-case (biphasic tablet and biphasic PK-Profiles):

❝ peak 1 (first tmax) around 2 hours, peak 2 about 6 hours, local minimum about 4 hours.

❝

❝ First, there are two possible cut-off time points:

❝ - 2 h (as early exposure - FDA)

Doesn’t makes sense according to the GL (and to me as well).

❝ - 4 h (for separating the two phases as shown in the profile)

❝ In my Opinion the 4 h as cutoff time-point is the right one, because it is - in contrast to an 2 h cut-off time point - possible to determine the first Cmax appropriately.

Yep, if (if!) you can distinguish between the phases.

❝ Second, there are two different ways to define the Cmax of the phases (cutoff is 4 h, next blood sampling time point 4.5h) in accordance to the guideline :

❝ - Cmax(0-4h) and Cmax(4h - t)

❝ - Cmax(0-4h) and Cmax (4.5h - t)

❝

❝ If the profiles are 'as expected', the results are of the two approaches are the same. But if there are some 'not expected' profiles (for example only one peak), Cmax(0-4h) and Cmax(4h - t)could be the same (4h-)value.

❝

❝ From scientific point of view, i would prefer: Cmax(0-4h) and Cmax (4.5h - t). But in studies i would prefer the other approach.

I can only tell you what I use. Set the cut-off to 4 hours. Calculate pAUC_0-4 and pAUC_0-t. Note that I use the actual times. Say you have samples at 3.5, 4.083, and 4.5 h (the 4 h sample five minutes late). Phoe­nix/Win­Non­lin will interpolate between 3.5–4 and between 4–4.083. If you use the lin-up/log-down trape­zoi­dal, no big deal. For an example see this post.
The C_max-values are a little bit tricky. I set up a filter based on actual time points and split the dataset. All samples with t ≤4 h go to the first part and all with t >4 h to the second. I calculate the two C_max-values sepa­ra­te­ly, rename them to sumfink meaningful and merge the results later. Note that if you have time devi­ations around the cut-off the number of samples / subject in each part must not be the same (say 6/10 in one subject and 7/9 in another).