Bioequivalence and Bioavailability Forum

Dear Dr. Shiv!

❝ There is no separate definition for differentiating compartmental and non-compartmental models specifically for WinNonlin.

Oh yes, it is:

❝ For beginners Handbook in Pharmacokinetics by WA Ritchel is a good ready reference book.

Fully agree. Since Wolfgang Ritschel was born and studied in Austria (he moved to the USA in 1968) – as an Austrian myself – I have to insist on the correct spelling: Ritschel – without the letter ‘s’ the name would be almost impossible to pronounce in German…

❝ Briefly, Compartmental models do not assume instantaneous distribution of drug throughout the body and you may find two or more segments in the elimination phase when you plot log concentration versus time on a graph paper.

Now I have to disagree with you. Such a behaviour is exactly what we expect from a 1-compartment model, i.v. administration.

❝ Whereas in non-compartmental model you will find a straight line in the elimination phase, this is because the drug is assumed to be instantaneously distributed througout the body and the drug is mostly confined to blood volume. These compartments can be explained using macro or micro rate constants. I think WinNonlin has both the options.

OK, now you are mixing two things up.
NCA claims to be model-independent, decribing the profile only by parameters directly accessible (i.e., without curve-fitting) from the the curve – irrespectively of any underlying process, how simple or complicated this process ever might be.
To give you an example: We may have a formulation showing a lag-time in absorption, a two-segment absorption phase, enterohepatic recyling, and a ‘fast’ and ‘slow’ elimination phase – which renders modeling (at least if only plasma data are available) a nightmare - but we can still apply NCA (i.e., calculate AUC, MRT, C_max, t_max).

Micro- and macro-constants are applicable to compartmental models only.
Any multicompartmental model can be formulated as a sum of exponential terms:

\(C_{ij}=\text{A}_j\cdot \text{e}^{\text{B}_j\cdot t_i}\textsf{,}\)

where the index \(\small{i}\) denotes the sampling time point, and the index \(\small{j}\) the number of the exponential term.
The coefficients (\(\small{\text{A}_j}\)) and – in some terminologies only the exponents – (\(\small{\text{B}_j}\)) are called macro-constants.

If you are building your own model from differential equations, you end up with an amount (the dose), volumes of distribution for all compartments, and transfer rate constants (in the jargon of modelers: micro-constants).
Theoretically it’s possible to calculate analytically (i.e., exact in the mathematical sense) micro- from macro, and vice versa. For fairly complicated models these solutions are give in PK textbooks (e.g., JG Wagner, Pharmacokinetics for the Pharmaceutical Scientist), but if your model is not given there, you should take a course in Laplace-transformations first…

WinNonlin has the option the formulate the model in both ways; it’s just an option which estimated constants are given first – the respective other ones are also given for all models.

BTW: Please don’t full quote – see the policy and TOFU; the preferred method is an inline reply. Otherwise moderators have to edit all your posts…