Dear all:Back to the Top
We are trying to use Bayesian forecasting to predict
Lopinavir (anti-retroviral) concentrations in
HIV-infected patients on salvage therapy. Lopinavir
is given orally every 12 hours. We plan to do so by
using a group of PK profiles to formulate the
sub-population PK parameters, then use a Bayesian
forecast program to test (and to predict) prediction
of drug concentrations. Here is the problem:
We are using WinNonlin version 4.1 to model our
Lopinavir PK data. Since all our patients are on
steady-state and the PK profiles are never "perfect"
looking, we use non-compartmental modeling for our
profiles. Then we plan to use the PKS (Abbottbase)
Bayesian program to make the predictions. However,
when setting up the new drug in the PKS Bayesian
program, we are asked to choose a compartmental model
for the drug and enter parameters such as absorption
rate (Ka), which we CANNOT obtain (directly) from the
non-compartmental modeling from WinNonlin. I have
been trying to calculate such parameters from our
non-compartmental analyses. The PKS program does not
seem to accept parameters from non-compartment
analysis.
We have also tried modeling our PK profiles with
WinNonlin using compartmental analysis, but it was
unsuccessful. I wonder if we need to resort to use
parameters from other literature, which may not be
suitable for our subgroup analysis.
I am wondering if there are other Bayesian programs
that would take non-compartmental parameters since our
PKS program is very old. We would appreciate any
insight into new Bayesian programs that could help us
do this analysis. Any other advice on how we should
tackle this program would be greatly appreciated.
Thank you!
Regards,
Lillian Ting
University of British Columbia
Yes. You might use the USC*PACK software. It makes both parametric, andBack to the Top
better yet, nonparametric population models, and does not only Bayesian
forecasting, but proper Bayesian adaptive control. Info, and a demo
version, are at www.lapk.org. Also there are a number of other bits of
material, and a variety of teaching topics.
Very best regards,
Roger Jelliffe
Roger W. Jelliffe, M.D. Professor of Medicine,
Division of Geriatric Medicine,
Laboratory of Applied Pharmacokinetics,
USC Keck School of Medicine
2250 Alcazar St, Los Angeles CA 90033, USA
Phone (323)442-1300, fax (323)442-1302, email= jelliffe.at.usc.edu
Our web site= http://www.lapk.org
Yes, you can use the USC*PACK collection programs ofBack to the Top
Dr. Roger W. Jelliffe's PK Lab. The prgrams for MAP
Bayesian forecastning, both parametric and
non-parametric, work well and give consistent results.
I strongly recommend you to contact Dr. Roger W.
Jelliffe for further details and advices.
My personal experience is very promising.
Kind Regards,
D. Terziivanov
Dimiter Terziivanov, MD,PhD,DSc, Professor
Head, Department of Clinical Pharmacology and
Pharmacokinetics,
Clinic for Therapeutics and Clinical Pharmacology,
Univ Hosp "St. I.Rilsky",
15 Acad. I. Geshov st, 1431 Sofia, Bulgaria
Tel:(+ 359 2)8510639;(+ 359 2)5812 828.
Fax:(+ 359 2)8519309. e-mal: terziiv.aaa.yahoo.com
Dear Lillian,Back to the Top
If you want to predict drug concentrations, you need a pharmacokinetic
model describing the concentrations. This is usually done by
compartmental modeling, but in theory any model that describes the
concentration pattern adequately might be used. Noncompartmental
modeling does not describe the complete concentration - time profile;
it 'summarizes' the profile in a few numbers, e.g. MRT. As far as I
know there is no way to convert MRT to a compartmental model, except
for the most simple case, i.e. intravenous administration and a
one-compartment model.
I would suggest to develop a pharmacokinetic model from your data by
appropriate population analysis. Then you can use this model and
parameters for Bayesian forecasting to predict concentrations in
patients and to adjust the dose accordingly. Steady state is no problem
for such an analysis. 'Perfect' data are unknown in the real world, and
one has to deal with this by an appropriate analysis. Of course, if
your data are really 'bad', you will not get a satisfactory result,
irrespective of the method you use. Noncompartmental analysis is not a
trick to deal with 'bad' data (although I am afraid it is used for this
purpose ...).
You stated that modeling your PK profiles with WinNonlin using
compartmental analysis was unsuccessful. Did you use a population
approach? In particular in case of 'poor data' (low number of
measurements per patient) and 'bad data' (unexpected profiles) a
population analysis is far superior to an individual analysis of each
patient. Of course it cannot correct erroneous data, but you will get
at least plausible parameters.
If you cannot obtain reliable PK parameters from your data, you could
use parameters from literature. Of course these may be less suitable
for your subgroup analysis, but in a data-poor situation it is better
to use all available data in a Bayesian approach.
The program MW\Pharm is able to perform a Bayesian population analysis
of a set of patient data, and to do a Bayesian forecasting, all within
the same program. The current version is a DOS-program (running well
under Windows), and a Windows version will be released towards the end
of 2004.
Information how to get the program from Mediware can be obtained by
sending an e-mail to: info.-a-.mediware.nl
If you have technical questios with respect to the program MW\Pharm,
you can contact me.
Best regards,
Hans Proost
Johannes H. Proost
Dept. of Pharmacokinetics and Drug Delivery
University Centre for Pharmacy
Antonius Deusinglaan 1
9713 AV Groningen, The Netherlands
tel. 31-50 363 3292
fax 31-50 363 3247
Email: j.h.proost.-at-.farm.rug.nl
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)