Back to the Top
Lately, there has been alot of traffic in this discussion group about
compartmental models, i.e., how multiexponential fitting relates to
compartmental models, the advantages of compartmental modeling, its
disadvantages, etc.
In general, compartmental models get a lot of bad press and kineticists
are reluctant to use them in drug development. People talk about the
issue of the number of compartments, non-identifiability, trying to
intrepret parameter constants, etc and because of this, most
pharmacokinetic analyses in drug development center on non-compartmental
analyses assuming that non-compartmental models are model-independent (a
wrong assumption).
Population pharmacokinetics has the same disadvantages of classical
compartmental models (assumption of a specific model which may be
non-identifiability) in addition to others that are often not discussed
(collinearity among predictor variables being the most egregious). Yet
no one questions a pop pk analysis to the same extent as classical
compartmental modeling, even though pop pk analyses often have fewer
samples per subject albeit with greater numbers of subjects. It is as
though having greater numbers of subjects gives a pop pk model greater
validity than doing compartmental analysis on a data-rich data set with
fewer subjects.
Any comments on this? Just food for though for the AAPS focus group
next month.
Peter L. Bonate, Ph.D.
Hoechst Marion Roussel
Clinical Pharmacokinetics
P.O. Box 9627 (F4-M3112)
Kansas City, MO 64134
fax: 816-966-6999
phone: 816-966-3723
Back to the Top
I really don't think there is much difference in philosophy among
"classic" (multi-sampling) and population (sparse sampling)
approaches in the caution that needs to be exercised against applying
the "wrong" model (with or without covariates). If the compartmental
model is inadequate, the intrapatient variance (SIGMA) estimated in
parametric population modeling is greatly inflated (although other
sources obviously contribute such as assay error, inaccurate dosing
and sampling times, etc). For ours, when we do population modelling
we are very aware of the pitfalls of implementing an inappropiate
model, e.g. when there area total of 2 or 3 samples taken at
different post-dosing times perhaps over several dosing intervals.
We usually screen 1 and 2 (or, occasionally, 3) compartment models
and examine weighted residuals and parameter estimates and SEs of
estimation. The latter is important when looking at possible
multicollinearity of covariates. If an appropariate model-building
strategy is used, e.g. individual screening of covariates,
stepwise addition of "significant" (P<0.01?) factors followed by
backwards elimination, then the possibilty of variables "masking" one
another is greatly reduced (but it can still happen). We have no
problem with the use of compartment models, as compared with other
representations of the data since they do have a degree of
physiological relevance. In NONMEM estimates of the more relevant
primary parameters of interest (clearances and volumes) can be sought
directly by reparameterising those dreadful compartmental rate
constanrts.
Cheers,
BC
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Bruce CHARLES, PhD
School of Pharmacy
The University of Queensland
Brisbane, Qld, Australia 4072
Telephone : +61 7 336 53194
Facsimile : +61 7 336 51688
Email : Bruce.Charles.at.pharmacy.uq.edu.au
+ + + + + + + + + + + + + + + + + + + + + + + + + + + + +
Back to the Top
On Mon, 20 Oct 1997 12:26:55 -0500 "Bonate, Peter, HMR/US"wrote:
> In general, compartmental models get a lot of bad press and kineticists
> are reluctant to use them in drug development.
> Population pharmacokinetics has the same disadvantages of classical
> compartmental models (assumption of a specific model which may be
> non-identifiability) in addition to others that are often not discussed
> (collinearity among predictor variables being the most egregious). Yet
> no one questions a pop pk analysis to the same extent
IMHO those who discard compartmental approaches have a
different philosophy of where the real problems lie. It
seems to me that they think there is some kind of "true"
and "objective" answer uncontaminated by assumptions which
can be achieved by invoking the "non-compartmental" mantra.
Population modellers have sidestepped/overcome this
particular intellectual hurdle and seem to find additional
insight from applying models (whether individual or
population) and can live with the assumptions and
uncertainties.
I can assure you that the population modellers I know
frequently discuss the problems with the technology and are
continually seeking better solutions. From a limited
perspective as an author, referee and editor of papers
dealing with population models I would say that population
models come in for a lot of peer criticism during the
manuscript review process. This is valuable for authors and
referees but readers of published articles may not
appreciate what has gone on before the MS appears in print.
--
Nick Holford, Dept Pharmacology & Clinical Pharmacology
University of Auckland, Private Bag 92019, Auckland, New Zealand
email:n.holford.-a-.auckland.ac.nz tel:+64(9)373-7599x6730 fax:373-7556
http://www.phm.auckland.ac.nz/Staff/NHolford/nholford.htm
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)