Back to the Top
I have been given some PK data to analyze. It is provided as concentrations
at given time points. The data is such that I only have a single measurement
for each animal at each time. So for example I have three concentrations
measurements each one from a different animal. So now I have been asked to
do some AUC analysis. What is the best way to do this? Can I take the data
and average it and plot it and calculate an AUC and somehow get and SD. Or
should I plot each individual animal and then average the AUCs that I come up
with from the three animals? It's been a while since I did this kind of
stuff as I have been doing computer stuff for a few years. Thanks in advance
for any help.
Jim Liddil M.S.
Manager of Information and Technology Services
Phytoceutica Inc.
(203)781-2544
jliddil.aaa.phytoceutica.com
Back to the Top
[Three replies - db]
From: martin.schumacher.at.pharma.novartis.com
Date: Tue, 10 Sep 2002 19:40:09 +0200
To: david.-at-.boomer.org
Subject: Re: PharmPK PK/AUC analysis
Dear Jim,
I suggest that you calculate the AUC for each animal separately and them
summarize them by their geometric mean. The variance of the geometric mean
can be calculated according to the formula in the contribution of J.
Attermann to this list on Sept. 8th.
Best regards,
Martin
Martin M. Schumacher, Ph.D.
Novartis Pharma AG
Basel, Switzerland
---
From: Paul Hutson
Date: Tue, 10 Sep 2002 13:08:34 -0500
To: david.aaa.boomer.org
Subject: Re: PharmPK PK/AUC analysis
Jim:
This sounds like a good task for NONMEM population analysis; can't
comment on Kinetica or WinNonLin, or USCPACK, but they may be other
options.
Paul
---
From: "Joseph Balthasar"
Date: Tue, 10 Sep 2002 16:04:22 -0400
To: david.-at-.boomer.org
Subject: RE: PharmPK PK/AUC analysis
Dear Jim:
One nice approach for such data is presented in:
Nedelman J. R., Gibiansky E.,Lau D. T., Applying Bailer's method for AUC
confidence intervals to sparse sampling, Pharm Res 12: 124-128, 1995.
Good luck!
J. Balthasar
Back to the Top
Jim
It appears from your description of the experimental design that you are
working with rodents. Calculate the mean and SD of the concentration values
at each time point. Estimate the AUC by a linear or log linear trapezoidal
method from the mean concentrations. Estimate the SD of the AUC by
consulting the equation in the following publication
Estimation of Variance for AUC by Jinhua Yuan in J. Pharm Sci, Vol. 82 (7),
761-763, 1993
Good Luck
Anup Zutshi Ph.D.
Global Drug Metabolism
Pharmacia
Mail Stop: 7260-300-104.1
Kalamazoo, MI 49001
Anup.Zutshi.-at-.Pharmacia.com
Back to the Top
Anup,
I think this approach is ill-advised; the resulting profile will be
non-representative
of any of the individuals in the study and the AUC may not be a good
estimate of the mean. In my view, the best solution would be to set
up the PopKin model in NONMEM, simulate an appropriate number of
individual profiles and calculate the AUC's individually.
Then, calculate the median and the 95% limits, alternatively the
median and the CV of the logged AUCs, if indicated.
The two-stage approach suggested by Martin Schumacher may give a fair
estimate if the number of individuals is high, however 3 individuals
sounds a bit scarce for this, in my view.
BR
Thomas Klitgaard =20
Pharmacometrican Clinical Pharmacology
Novo Nordisk A/S
Krogsh=F8jvej 51, 9ES.35
2880 Bagsvaerd
Denmark
tkt.-a-.novonordisk.com www.novonordisk.com =
Back to the Top
Tom
From my understanding of Jim Liddl's original question, he had a design
where at each timepoint 3 animals had been bled, however each animal
did not
yield a complete PK profile. So in essence there were more than 3
animals (3
animals X # of timepoints taken)that represented a single PK profile. No
individual animal in itself yielded a complete profile. Although this
design
can be analyzed by a Popkin model using Nonmem or any other suitable
software (no argument there), it is probably more than he is
interested in
analyzing since all he wanted was to estimate an AUC. In lieu of that
the
suggestion that I had made seems appropriate and adequate.
Anup
Back to the Top
This is correct. The experiment was setup as such:
Sample Time
Animal 1,2,3 5 min and 90 min
Animal 4,5,6 15 min and 2 hr
Animal 7,8,9 30 min and 4 hr
Animal 10,11,12 45 min and 8 hr
Animal 13,14,15 60min and 24 hr
I had not input on how to setup the experiment. I have only been asked to
analyze the data. Apparently the internal standard technique was used and
each sample was only extracted once. I can hear the groans folks must be
making. And I do realize this is a case of Lies, D.at.mn Lies and Statistics.
Jim Liddil
PharmPK Discussion List Archive Index page
Copyright 1995-2010 David W. A. Bourne (david@boomer.org)