Back to the Top
Hello there,
I have some friends and colleagues who are interested in PK modeling and wanted
to experiment for themselves, rather than just looking over my shoulder. Since
this is "for fun" their budget is negligible so I was hoping to point them to
free (or near-free) programs (this includes avoiding the need for e.g. a
commercial Fortran compiler). I have some past experience with SAAM (as
WinSAAM) and think it might be a good starting point. However, since I may have
to fend off challenging questions I wanted to improve my documentation library.
At the moment I rely on the book (Wastney et al.), the WinSAAM help files, and
example model description files I have accumulated over time. The WinSAAM help
facility is enough to get started but doesn't cover all of the SAAM commands, or
some of the more arcane options. The canonical document appears to be the SAAM
User's Guide and related material but I don't see that available anywhere.
Does anyone know if the SAAM manuals are available anywhere?
Any other suggestions? I am looking at Boomer, but don't yet have enough
familiarity to be of much help. Is Monolix still free now that it is from
Lixoft?
Thanks for any suggestions.
All the very best,
Bernard
Bernard Murray, Ph.D.
Senior Research Scientist, Drug Metabolism
Gilead Sciences, Foster City CA
Back to the Top
Bernard,
There is book available at Amazon that explains modeling with SAAM at the
primary tool used in the examples. Not a manual of course. And it is old,
1998.
Investigating Biological Systems Using Modeling: Strategies and Software
[Hardcover]
Meryl E. Wastney (Author), Blossom H. Patterson (Author), Oscar A. Linares
(Author), Peter C. Greif (Author), Raymond C. Boston (Author)
Investigating Biological Systems Using Modeling describes how to apply software
to analyze and interpret data from biological systems. It is written for
students and investigators in lay person's terms, and will be a useful reference
book and textbook on mathematical modeling in the design and interpretation of
kinetic studies of biological systems. It describes the mathematical techniques
of modeling and kinetic theory, and focuses on practical examples of analyzing
data. The book also uses examples from the fields of physiology, biochemistry,
nutrition, agriculture, pharmacology, and medicine.
Key Features:
* Contains practical descriptions of how to analyze kinetic data
* Provides examples of how to develop and use models
* Describes several software packages including SAAM/CONSAM
* Includes software with working models
Dan Combs
Combs Consulting Service
[Probably the old mainframe SAAM-23, SAAM-25 but could be useful. The SAAM II
website might have some information.
http://www.saam.com/
I see some tutorials. The boomer program and manual is on www.boomer.org ;-) db]
Back to the Top
The following message was posted to: PharmPK
Thanks for the comments I received so far. I already have the book that Dan suggested
(that's the one I mentioned in my original post) and it is a great start but not fully
comprehensive.
The saam.com website is for SAAM II (the "U Washington branch" of SAAM), which, while
being descended from the original SAAM/CONSAM, has been heavily re-engineered and uses a
graphical interface. WinSAAM (the "U Penn branch"?) is more close related to the original
SAAM and still uses textual model description files. This means the SAAM II documentation
at saam.com isn't much help for SAAM/WinSAAM.
Sounds like I should invest some time playing with Boomer...
All the very best,
Bernard
Bernard Murray, Ph.D.
Senior Research Scientist, Drug Metabolism,
Gilead Sciences, Foster City CA
Back to the Top
I spent a couple of months in 2005 working with Dr Ray Boston (U Penn) trying to get to
grips with WinSAAM and would not recommend it for PK modelling because of its quite unique
nomenclature and non-obvious methods. Any effort put into learning PK modelling with
WinSAAM would be lost when trying to deal with, and especially report, PK problems.
Modern PK modelling software is available for free to academic institutions (Pharsight
Phoenix nlme, Lixsoft Monolix).
Nick
[I think Nick is talking about the old SAAM. SAAM II is a much newer and more current
piece of PK software.- db]
Back to the Top
The following message was posted to: PharmPK
And this website is still up and running.
http://www.saam.com/support/index.html
Essentially the entire user's manual as separate .pdf chapters.
Dan Combs
Combs Consulting Service
Back to the Top
The following message was posted to: PharmPK
[Dan Combs, the documentation on the saam.com website is for SAAM II (not SAAM),
which is rather different from SAAM and WinSAAM.]
Nick and others, thanks for the reminder about academic pricing. If any of the
licenses allow "non-commercial use" (e.g. for self-study at home), rather than
needing the user to be associated with an academic institution, then that would
work.
Actually, I'd disagree with Nick's notion that WinSAAM (and classic SAAM) should
be avoided. The nomenclature isn't that unique (except perhaps for the use of
field-based text in the model description files). It is not that difficult to
make the mental switch between e.g. k12 and L(2,1). I rather like the way that
SAAM is smart enough that you don't have to explicitly define compartments or to
write differential equations (they are inferred from the existence of the rate
constants). I use NONLIN, NONMEM, Berkeley Madonna and SAAM II on a fairly
frequent basis but haven't discarded WinSAAM as an option. I agree with Nick
that some of the methods take some experimentation before the "Aha!" moment
(e.g. delay methods, forcing functions and time control) and the EMSA popPK
function probably looks a little dated, but there are often reasonable example
files available (e.g the library now at Purdue
http://www.cfs.purdue.edu/fn/model/ ). My only wish is that the program
documentation was more comprehensive (HELP doesn't work in the WinSAAM command
window, and the Windows-based help files are spotty). I don't think people who
learned modeling using WinSAAM would have difficulty making the transition to
more mainstream programs.
One other free suggestion I received was to try Gepasi ( http://www.gepasi.org/
).
All the very best,
Bernard
Bernard Murray, Ph.D.
Senior Research Scientist, Drug Metabolism,
Gilead Sciences, Foster City CA
Back to the Top
The following message was posted to: PharmPK
Dear Nick and All:
If you are really interested in good population PK/PD modeling, you
might get away from the constraints of parametric approaches to deal with
the problem. Nonparametric approaches are not only unconstrained by any
parametric assumptions concerning the shape of the parameter distributions
(permitting more likely results given the data), but they also lend
themselves to the design of maximally precise dosage regimens, which hit
desired therapeutic targets with minimal expected weighted squared error.
Dosage regimens developed using single valued parameter estimates do
not hit targets most precisely, as they are limited by the separation
principle. This principle states that whenever one seeks to control a system
(design a dosage regimen to hit a target) by first getting single valued
model parameter estimates, and then using these estimates to control the
system, the task is done suboptimally, as there is no performance index to
optimize the process. One simply develops a regimen to hit the target
exactly, and we all know that it does not. There is no tool there to
evaluate and optimize the expected precision with which the target will be
hit [1].
In contrast to parametric approaches, which estimate central
tendencies of the parameter distributions, and then estimate the random
variability about those fixed effects, the nonparametric approach proceeds
quite differently. It employs the theorems of Caratheodory, Lindsay, and
Mallet, which show that one need not search over the infinite continuum of
all possible parameter distributions. Instead, the most likely distribution,
given the data, can be found in a discrete collection of support points, up
to 1 for each subject studied. No assumptions are made at all about the
shape of these distributions - they are determined only by the data itself.
The results they obtain are more likely than those obtained using parametric
approaches, as they are not constrained by any of the parametric assumptions
made about the shape of the model parameter distributions [2]. Evan Paul
Baverel showed that the nonparametric option developed for NONMEM was better
than parametric approaches [3], even though it was far from rigorous or
optimal [4].
Further, if one really wishes to use population PK/PD models for
something socially useful like developing dosage regimens to optimize care
for individual patients, the nonparametric approaches are most naturally
linked to the process of multiple model (MM) dosage design. This comes, not
from the PK community, but from the aerospace community, where it has been
quite widely used for flight control and spacecraft guidance systems. Almost
all commercial airliners and fighters today use MM Bayesian adaptive control
as the tool for controlling the behavior of these craft. You do not fly
planes using modeling, simulation, and Bayesian forecasting! It takes
maximally precise MM Bayesian adaptive control.
Nick, you once said to me that you would not believe the
nonparametric approach until you read about it in a peer reviewed journal.
First, you may not have read the right journals [5-7], but now the evidence
is all around you. Please, Nick, I most respectfully ask, nay implore, you
to consider that there can be better ways to do pop modeling than NONMEM or
other parametric methods! One easily available comparison of the
nonparametric approach versus parametric ones is in [2].
For those of you who are interested in the nonparametric approach to
population PK/PD modeling, the new Pmetrics package, embedded in R,
developed by Michael Neely MD in our lab, is readily and freely available by
contacting him at mneely.at.usc.edu, or from our web site www.lapk.org.
Very best regards,
Roger Jelliffe
References:
1. Bertsekas D: Dynamic Programming: deterministic and stochastic
models. Englewood Cliffs (NJ): Prentice-Hall, 1987; pp.144-146.
2. Bustad A, Terziivanov D, Leary R, Port R, Schumitzky A, and Jelliffe
R: Parametric and Nonparametric Population Methods: Their Comparative
Performance in Analysing a Clinical Data Set and Two Monte Carlo Simulation
Studies. Clin. Pharmacokinet., 45: 365-383, 2006.
3. Baverel P: Development and Evaluation of Nonparametric Mixed Effects
Models. Ph.D. Thesis, University of Uppsala, ISBN 978-91-554- 7995-4,
Department of Pharmaceutical Biosciences, Uppsala University, Uppsala,
Sweden, 2011.
4. Leary R and Chittenden J: A Nonparametric Analogue to POSTHOC
Estimates for Exploratory Data Analysis. A poster presentation at the
Population Approach Group Europe meetings in Marseille, France, June 18-20,
2008.
5. Lindsay B: The geometry of mixture likelihoods: A general theory.
Ann. Statist. 11: 86-94, 1983.
6. Mallet A: A maximum likelihood estimation method for random
coefficient regression models. Biometrika, 73: 645-656, 1986.
7. Schumitzky A: Nonparametric EM Algorithms for Estimating Prior
Distributions. App. Math. and Computat. 45: 143-157, 1991.
Roger W. Jelliffe, M.D., F.C.P.
Professor of Medicine,
Co-Director, Laboratory of Applied Pharmacokinetics
www.lapk.org
USC Keck School of Medicine
2250 Alcazar St, Room 134-B
Los Angeles CA 90033
|
Want to post a follow-up message on this topic?
If this link does not work with your browser send a follow-up message to PharmPK@boomer.org with "Availability of manual for SAAM" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)