Back to the Top
Perhaps Nick could recognise the difference between the ivory towers of
academe and the real world? In the real world most major pharmaceutical
companies insist on simple, non-compartmental analysis of toxicology
studies. Such studies rarely, if ever, contain sufficient sample points to
adequately describe a model of more than 1 compartment, primarily on cost
and animal welfare grounds.
Yes, in theory, one can use the population pk approach for toxicology
studies in rodents and I know of a number of references describing this
approach. However, as I said, in the real world, for the majority of tox
studies, pharmaceutical companies insist on NCA.
HUNTINGDON LIFE.SCIENCES LIMITED
[Hmmm. This could be interesting!
As most of you know I (db) am also an academic...well I try to be
anyway. Also, I do prefer to model data as I believe it is the best
way to get the most information out of any set of data. I also
recognize that modeling can be quite (expensive) labor intensive and
Non-compartmental analysis can be quick and automated easily. It can
be very useful (I gather ;-)) if it is done properly. [Reminds me of
message posted to PharmPK recently - can AUC(0-last) be greater than
AUC(all). I'm sure that we have all seen data sets where the
concentrations at later (and other times) rise when they 'should'
fall due to random assay (etc.) error. If we use the last two points
to estimate terminal slope this may well be positive and give a
negative value for the k(z) and the Cp(last)/k(z) values. Of course
any review of the data by person or software should catch this
Sorry Alan, but I'm not impressed by the lack of a number of
sufficient data points argument (that is what Bayesian or population
PK analysis for). It reminds me a conversation with a colleague many
years ago. This colleague was trying to argue for AUC calculation
over modeling because they only had 2 or 3 points AND there was
considerable error in the data. I'd hate to think what a 'full' SHAM
(SHape, Area, Moment) analysis would produce with these results. I
don't have as much problem with AUC calculated to estimate drug
exposure with few points if the limitations are fully realized.
Non-compartmental analysis and/or modeling done properly can give
For balance I should mention a very old set of data I was asked to
look at when I was starting out in PK analysis. The assay was by
radioactivity determination after paper/glass slide chromatography.
The original analysis included a model with step functions and
multiple exponential terms turning on and off after a single dose
exposure. It sure was a fancy model produced I should say by an
industrial scientist (maybe newly qualified by academia ;-)
Good data always makes the subsequent analysis easier.
This could be an interesting thread. I wasn't sure about the subject
line but it is probably as good as any. Let the Games begin - db
(aka David Bourne)]
Back to the Top
This will be an interesting thread. I will put my two cents in and agree
with Alan with regard to non-clinical toxicokinetics studies and the utility
of a pop PK approach. While I enjoy modeling (gasp! I can't believe I would
ever say that), I do have to agree that in the real world of drug
development, there simply isn't a real reason to go beyond what simple
determinations you want from your non-clinical animal data to allow you to
proceed in the clinical trials. Building a nice population PK model in rats
in a nice exercise, but when it comes to obtaining marketing approval, the
emphasis is on the proper characterization of the drug in humans.
Jeffrey L. Larson, Ph.D.
Director of Toxicology and Pharmacokinetics
4888 Loop Central Drive
Houston, TX 77081-2225
Back to the Top
[Two replies - db]
From: "David S. Farrier"
Date: Fri, 28 Jun 2002 14:24:26 -0600
Subject: Re: PharmPK Walt Disney School of Pharmacokinetics
As an aside to this budding topic on academic, real world, or Mickey
Mouse pharmacokinetics, for those interested in the "NCA software
that pharmaceutical companies insist on" [paraphrasing Alan], take a
look at PK SOLUTIONS at www.SummitPK.com
This Excel-based program computes some 75 noncompartmental parameters
with a few clicks of the mouse (the button, not the animal). And the
results, where appropriate, agree precisely with those obtained by
more complex compartmental modeling software.
While visiting the site, download a listing of the 75 equations for
your own use.
David S. Farrier, Ph.D.
Summit Research Services
Montrose, Colorado 81401
From: Volosov Andrew
Date: Fri, 28 Jun 2002 22:41:02 +0200
Subject: RE: PharmPK Walt Disney School of Pharmacokinetics
This is a good one, no question.
I believe one should always LOOK at the data, before starting to play with
Looking at the data that started this discussion, it is clear that the Tmax
is somewhere between 1h and 6h. AND there were only two samples taken
between these timepoints. This may sound provocative (in line with the rest
of this thread :)), but IMHO, saying that "the Tmax is between 1h and 6h" is
at least as good as forcing it to be at 4h just because you happened to take
a sample at 4h timepoint and the concentrations you measured there were
slightly higher than at 2h. If you need anything better than that, there's
no way around using a model. What kind of a model is a whole different
story. If you are lucky and your data fits kit-kat equation (which is quite
often the case, despite the fact that it belongs to Mickey Mouse
One-Compartmental world), you might be able to use the approach we published
back in 1999 (Volosov A., Bialer M. Biopharm Drug Dispos 1999
Jan;20(1):3-9). It uses the MRT and portions of AUC before and after the MRT
to calculate Tmax. Worked very well for a carbamazepine study where we
clearly missed the Tmax because of the sampling regimen.
Curious to see more replies.
Andrew Volosov, PhD
UCB Research Inc.
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (firstname.lastname@example.org)