Back to the Top
My group and I are doing curve fits to effect data for neuromuscular =
blockers. (example: Pharmacokinetic analysis of the effect of vecuronium =
in surgical patients, Anesthesiology, 88:874, 1998). We are still using =
MKModel a DOS program from Biosoft (Cambridge) which is pretty clunkey =
by to-day's standards and very demanding of the original parameter =
estimates. I would like to upgrade to something a little easier to use. =
It needs to be programmable for user defined functions, to use various =
models (compartmental, non-compartmental etc..) and handle the fitting =
to at least 2 effects.
We know of 2 programs which will do this, NONMEM but this would require =
Fortran to program and I don't relish learning this or buying a =
compatible compiler. The second is ModKine, expensive and unknown to us.
We would be grateful to hear from people who have done such work and =
have experience with either of these programs, or know of others which =
we should consider.
With thanks,
Tom Torda
Back to the Top
The following message was posted to: PharmPK
Tom,
"Thomas A Torda (by way of David_Bourne)" wrote:
>
> PharmPK - Discussions about Pharmacokinetics
> Pharmacodynamics and related topics
>
> My group and I are doing curve fits to effect data for neuromuscular =
> blockers. (example: Pharmacokinetic analysis of the effect of vecuronium =
> in surgical patients, Anesthesiology, 88:874, 1998). We are still using =
> MKModel a DOS program from Biosoft (Cambridge) which is pretty clunkey =
> by to-day's standards and very demanding of the original parameter =
> estimates. I would like to upgrade to something a little easier to use. =
I fully agree. MKMODEL is past its use by date and many other
programs are easier to use.
I hardly use it anymore. There is only one thing it can do that
NONMEM cannot -- solve implicit equations (note NONMEM claims to do
this but I have never been able to get it to work satisfactorily).
This is essential to use such things as isobole models but this is
not a common task.
> It needs to be programmable for user defined functions, to use various =
> models (compartmental, non-compartmental etc..) and handle the fitting =
> to at least 2 effects.
> We know of 2 programs which will do this, NONMEM but this would require =
> Fortran to program and I don't relish learning this or buying a =
> compatible compiler.
I would recommend learning how to use NONMEM. It does not require you
to learn FORTRAN (except for some minor dialect issues). The NM-TRAN
modelling language is straightforward. The G77 compiler is free and
IMHO works as well (if not better in some cases) than the Compaq
Visual Fortran compiler. Why not come along to the next PAWS/PAGANZ
meeting in Melbourne next January and take a crash course in how to
use NONMEM?
http://www.phm.auckland.ac.nz/Conferences/PAGANZ/
The beauty of NONMEM is its flexibility from simple individual curve
fits to full blown popln models. There is also a an active user group
and free consulting service for licensed users.
--
Nick Holford, Divn Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New Zealand
email:n.holford.aaa.auckland.ac.nz tel:+64(9)373-7599x6730 fax:373-7556
http://www.phm.auckland.ac.nz/Staff/NHolford/nholford.htm
Back to the Top
[Two replies - db]
From: "Hugh Barrett"
Date: Sun, 14 Oct 2001 14:05:08 +0800
To: david.at.boomer.org
Subject: Re: PharmPK Pk-Pd curve fitting
The following message was posted to: PharmPK
I would suggest you also look at the SAAM II program (www.saam.com). I am
using this program to look at methadone PK and its relationship with various
measures of PD. SAAM II is very flexible, has an intuitive graphical
interface and permits the user to define the relationship between PK and PD.
Direct and indirect effect models can be developed. Viewing the model
solutions is easy, as it obtaining the statistics associated with model fits
to the data. I suggest you have a look.
Hugh Barrett
Mailing address:
Department of Medicine, University of Western Australia
Box X2213 GPO, Perth, Australia 6847
Courier address:
Department of Medicine, MRF Building,
Rear 50 Murray St, Perth, Australia 6001
Phone 61 8 9224 0249 Fax 61 8 9224 0246
pbarrett.-at-.cyllene.uwa.edu.au
---
From: Roger Jelliffe
Date: Sun, 14 Oct 2001 16:28:07 -0700
To: david.-a-.boomer.org
Subject: Re: PharmPK Pk-Pd curve fitting
Dear Tom:
You might consider the USC*PACK software for population
modeling of large and nonlinear systems. There is a module (BOXES)
with which you draw your structural model. It generates the Fortran
source code for the differential equations, the output equations, and
the names of the various parameters. You can then either use the web
and get into our small 3-cpu Dell cluster or into the big parallel
machine at the San Diego Supercomputer Center, upload your model
file, and the results of running the preparation programs for the
IT2B or the NPEM or NPAG programs. The machine will then compile and
link everything. You do your analysis, and get the results emailed
back to you to see on your machine. The software currently handles
multiple outputs (your at least 2 effects). You can program it to do
almost anything you want. You might look at our web site
(www.lapk.org) and go to teaching topics, and look at parametric and
nonparametric population modeling. Information about the software and
its costs is below.
You might also wish to be able not just to model drug
behavior, but also to act most intelligently based on your results,
to develop the dosage regimen which is designed to achieve your
target goals with maximal precision. We are developing "multiple
model" dosage design methods to do this. Work to control 3
compartment linear models is close to completion. Work to control
larger and nonlinear models is well under way. There is also more
material about this new form of dosage regimen design on our web
site. I look forward to hearing more from you and discussing all this
more.
Very best regards,
Roger Jelliffe
Dear Colleague:
Thank you for your inquiry concerning our work. I am
enclosing some material describing our laboratory and its software,
the USC*PACK PC clinical programs, the IT2B and NPEM population
modeling programs, and the BOXES program for making large nonlinear
and effect pharmacokinetic/dynamic models. I am also enclosing an
academic licensing agreement to use the programs.
The USC*PACK clinical programs employ a linear 3 compartment
pharmacokinetic model having an absorptive, a central (serum)
compartment, and a peripheral (nonserum) one. One can also enter and
store parameter values for any drug having this basic structural
model. There are population models available for clinical use to
guide and adjust therapy with gentamicin (general medical patients,
ICU patients, others, and several for newborns and premies). Similar
models are available for Tobramycin, Netilmicin, and Amikacin. There
are also models for digoxin, digoxin with quinidine, quinidine, and
digitoxin. The software also permits development of dosage regimens
to achieve target goals in the peripheral nonserum compartment as
well as to achieve target serum concentration goals. This is
especially useful for digoxin and digitoxin, where the main clinical
and toxic effects are correlated much better with those peripheral
(tissue) concentrations that with serum concentrations.
Models are also available for lidocaine, theophylline
(several different ones for smokers, etc, and for long - acting
preparations), vancomycin, trimethoprim, and others.
In the user manual, also available for inspection and
printout on line at our web site, http://usc.edu/hsc/lab_apk/ are a
number of clinical case presentations (pages 40-50) which illustrate
the use of the programs in a variety of clinical situations.
In addition, there are about a dozen Technical Reports
available on our web site, on various PK/PD, clinical, and
mathematical aspects of pharmacokinetics, population PK/PD modeling,
and clinical applications. You can download them directly, read them,
and print them. For some files you will need the Adobe Acrobat
reader, which you can also get free from the web links to it.
The USC*PACK clinical programs also let you enter and store
population parameter values for any drug having the basic structural
model described above. They also employ MAP Bayesian fitting to make
individualized pharmacokinetic models, and they also compute AUC,
AUIC, AUC/MIC, % of dose interval above the MIC, peak/MIC ratios, etc.
The programs also compute diffusion into porous spherical objects
such as endocardial vegetations. These models can also be used to
simulate a post-antibiotic effect.
This input, as well as the serum concentration profile, can
then be used as input to effect models, of the growth and kill of
organisms, for example. In this way one can now evaluate the ability
of a certain dosage regimen to generate a serum level profile, and to
kill under defined circumstances, in addition to relating growth and
kill to the more common indices such as AUC, time above MIC, AUC/MIC,
etc.
The software for population pharmacokinetic modeling employs
both parametric (iterative 2-stage Bayesian, IT2B) and nonparametric
EM (NPEM) methods. Again, the basic structural model is the same.
Many menus are available for parameterizing this 3 compartment linear
model in many ways (rate constants, clearances, etc.).
Other much larger and quite nonlinear PK/PD population models
(any system which can be described by ordinary differential
equations) can now be made using the IT2B and NPEM approaches on the
Cray T3E and the new IBM Blue Horizon large parallel machines at the
San Diego Supercomputer Center. The Blue Horizon has 1152 processors,
and is the fastest academic machine available for nonclassified work.
Research accounts on the Blue Horizon can be obtained upon request
and approval. The BOXES program can be used to aid in preparing the
differential equations for these large and nonlinear models.
The software is made available by license from the
University. The requested donations for this version, whether or not
you are already a USC*PACK user, are:
Hospital or
Commercial
Academic Institution
or Industry
First donation: $395.00 $695.00
Upgrades thereafter: $295.00
$495.00
However, upgrading to the present version from version 10.6
or 10.7 is free.
We can handle checks or electronic bank transfers (ask us for
info), but not credit cards or purchase orders. Please sign and mail
the enclosed license agreement to use the programs back to us, along
with a check for the appropriate amount, made out to the University
of Southern California. We will send the programs back to you with
the user manual and other material as soon as possible. We supply the
programs on HD 3.5 inch floppy disks.
The programs run in Windows 95, 98, NT, or 2000 in the DOS
mode. They need about 610K free memory (the less memory-resident
software you have, the better), and need about 20 Mb free disk space
to install.
If you have any problems or questions at any time, please
write to the return address, call us at 323/442-1300, fax us at
323/442-1302, or email us at jelliffe.-a-.hsc.usc.edu. We are also on
the World Wide Web at http://www.usc.edu/hsc/lab_apk/. This site has
announcements of new events, new upgrades for some of the software
which you can download, and a number of publications and technical
reports.
Sincerely,
Roger Jelliffe, M.D.
Professor of Medicine
THE USC*PACK PHARMACOKINETIC PROGRAMS
WHAT'S NEW IN THE SETUP! CLIENT - SERVER MODES
The programs now run in 3 modes. The first is the usual mode
on a single machine. The second is on a network, in a classroom mode.
Each classroom machine accesses the programs from the network, but
keeps its patient data files in its own disk or directory. The third
is a true client-server mode. Here, all programs and patient data
files are kept on the central hospital server, for example, but are
accessed and run from the satellite (client) machines.
WHAT'S NEW WITH AMINOGLYCOSIDES
These programs have several new population PK models, for
general medical patients, ICU patients, young previously healthy
patients, for gentamicin in patients with spinal cord injuries, and
for newborn infants. As before, the GENTamicin, TOBramycin,
NETilmicin, and AMIKacin programs compute D-optimal times for
monitoring serum levels. Diffusion into vegetations and models of
bacterial growth vs killing at a stated MIC can also be evaluated as
described just below.
WHAT'S NEW WITH GENERAL MODELING, BAYESIAN FITTING
The General Modeling, Bayesian fitting program has an option
for entry of a chronic steady-state dosage regimen, followed by
subsequent changes in dosage or renal function. Other related
programs compute the dynamics of diffusion into a spherical simulated
endocardial vegetation, abscess, or a bacterium, nonlinear saturable
pharmacological effects, and the dynamics of bacterial growth and
kill, and the post-antibiotic effect (PAE). These help to evaluate
the efficacy of antibiotic regimens with respect to their ability to
kill adequately under various circumstances, for various values of an
organism's MIC. Models for simulating a PAE of 6 hours, and for
growth and kill of Pseudomonas by gentamicin, tobramycin, and
ticarcillin are now provided. You can also make and store your own
models.
WHAT'S NEW WITH THE IT2B AND NPEM POPULATION MODELING PROGRAMS
The IT2B and the NPEM programs have been very much overhauled
and upgraded. The NPEM program now automatically helps you select an
appropriate number of grid points to support your population joint
density. It has algorithms to make each computational cycle go
progressively faster. The 2D plots of the marginal and the 3D plots
of the joint marginal densities are enhanced. Scatterplots and
regression relationships are now available for predicted versus
measured serum levels in both the Iterative 2 Stage Bayesian (IT2B)
population modeler, from mean or median population parameter values.
Each patient's own Bayesian posterior parameter values can also be
used to predict the measured levels.
For the NPEM part, one similarly can predict measured serum
levels from population mean, median, or mode values, and can also
further compute each patient's individual Bayesian posterior
parameter joint density, and can see the predicted versus measured
serum levels based now on each patient's own mean, median, or mode
parameter values.
There are now also new versions of the IT2B and NPEM programs
that permit making large and nonlinear PK/PD models. These use a new
version of the BOXES program which greatly enhances the development
of the differential equations describing the behavior of the model,
the multiple output equations for the model, and the parameter names.
The user then accesses the Cray T3E or the new IBM parallel computer
at the San Diego Supercomputer Center, gets the relevant files from
his PC, does the analysis, and sends the result files back to his PC
to examine them.
THE OVERALL USC*PACK COLLECTION NOW INCLUDES:
1. The USC*PACK CLINICAL programs. These enhance precise
goal-oriented dosage design and therapeutic drug monitoring by
Bayesian individualization of drug dosage regimens. Drugs include
aminoglycosides, vancomycin, digoxin, digoxin with quinidine,
digitoxin, lidocaine, procainamide, theophylline, quinidine,
gentamicin in newborns, and TMP-SMX for PCP pneumonia. CCr is
computed between a single stable or pairs of unstable serum
creatinine levels. It can change from dose to dose, as can body
weight. Models of diffusion and bacterial growth and killing are
especially useful to evaluate adequacy of a regimen to kill under
stated conditions of growth rate, max kill rate, and MIC.
2. The USC*PACK IT2B and NPEM POPULATION pharmacokinetic
modeling programs employ an iterative Bayesian and a nonparametric
expectation maximization (NPEM) algorithm respectively. The strength
of the NPEM program is that it computes the entire joint probability
density for a 2 compartment absorptive model, or larger on the SDSC
Cray T3E, even with mixed oral and intravenous input, and thus can
obtain both Vd and bioavailability together. It makes no parametric
assumptions such as mean, standard deviation, etc., but also gets
them. It can thus discover unrecognized subpopulations. It reads
routine patient data files from the clinical programs. Parameter
values found can be entered and stored for use with the clinical
programs.
3. The USC*PACK BOXES program makes customized PK/PD models by
placing boxes on the screen and connecting them with arrows.
Equations are automatically written for the Model part of the ADAPT I
PC programs. Effect models (Hill or Keo) are easily made.
The programs run on the IBM PC and compatible machines. They
need at least DOS 5.0, and 16 MB of memory. The programs come on 3.5
inch disks. Please sign and mail the enclosed license agreement back
to us, with a check, made out to the University of Southern
California. We will send the programs as soon as possible. If you
have questions, write to us at 2250 Alcazar St., Los Angeles CA
90033, call us at 323/442-1300, fax us at 323/442-1302, or email us
at jelliffe.-at-.hsc.usc.edu. We are also on the World Wide Web at
http://www.usc.edu/hsc/lab_apk/ It has announcements of new events
and developments, and publications.
Sincerely,
Roger Jelliffe, M.D.
Professor of Medicine
The licensing agreement from USC to use the software is shown below.
LABORATORY OF APPLIED PHARMACOKINETICS
USC School of Medicine (CSC 134B), 2250 Alcazar St., LA, CA 90033
Phone (323) 442-1300, Fax (323) 442-1302
ACADEMIC LICENSE AGREEMENT (Version 9/12/98)
This Agreement is made between
("Researcher") and the University of Southern California
("University"), University Park, Los Angeles, California 90089, with
respect to computer software developed under the direction of Roger
W. Jelliffe, M.D., a U.S.C. faculty member, for use in studying drug
behavior and calculating probable drug dosage and infusion
requirements for patients and supporting documentation (both of which
are herein referred to as "Software"). The University has the sole
right to license use of the Software, and is willing to grant a
royalty-free license to the Researcher to use the Software for
research purposes under the following terms. The Researcher realizes
that the Software may be of significant commercial value to the
University, and Researcher desires to evaluate, test and use the
Software listed at the bottom of this document for research purposes
only.
1. This License is limited to use of the Software for
academic and research purposes by the researcher within the
institution designated below, and will not be disclosed, given or
sold to anyone outside of the institution. Duplication of the
Software for any purposes other than the above use, and for back-up
protection, is prohibited. The Software may not be used for any
commercial purpose without the express written agreement of the
University. Researcher will return Software and destroy all other
copies immediately upon request.
2. Researcher further acknowledges that this agreement will
also apply to all subsequent Software s/he may receive from Dr.
Jelliffe unless another written agreement is concluded. All Software
is and will remain the property of the University.
3. Researcher acknowledges that the Software is directed to
pharmacokinetic analysis and the control of drug dosage regimens.
Researcher understands that the University makes no warranties,
either express or implied, as to any matter concerning this Software,
including the condition of the Software, its usability, or fitness
for any particular purpose. Researcher agrees to hold the
University, its officers, employees and agents, including but not
limited to all individual developers of the Software, harmless for
any and all damages, expenses, claims or other liability suffered as
a result of Researcher's use of the Software. Researcher agrees that
s/he alone is responsible for whatever dosage regimen s/he or his/her
groups order for or administer to a patient.
4. This Agreement constitutes the entire Agreement between
the parties concerning the Software. No amendment or assignment
shall be binding on the parties unless mutually agreed to and
executed in writing by each of the parties. This Agreement shall be
interpreted in accordance with the laws of the United States and of
the State of California. Jurisdiction and venue shall lie with any
competent court within the County of Los Angeles, California. In the
event litigation or arbitration is commenced to enforce any of the
terms of this Agreement, the prevailing party shall have the right to
recover its reasonable attorneys' fees and costs of such litigation
or arbitration from the other party.
SOFTWARE LIST:
(The USC*PACK PC Clinical Collection) (The
USC*PACK Modeling Collection)
1. GENT 1.
MODEL, ID, ODE, SIM
2. TOB 2. MMID8, MMSIM8
3. NET 3. ID3,
SIM3, ID3CCR
4. AMIK 4. ID3P, SIM3P,
5. MB 5. BOXES
6. MLS 6. IT2B
and NPEM Population Programs
Date: __________Researcher (Please print or
type):____________________________________
Institution:________________________________________________________________________
Address:_________________________________________________________________________
City, State Zip
Code:____________________________________________________________________________
Country:_________________________
Phone:___________________________ Fax:_________________________
Email:______________________________
Please check Type of User: Hospital or Academic_______
Commercial or Industrial___________
Signature____________________________________________________
Date________________________
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)