Back to the Top
Hello All:
I am trying to do compartmental analysis of data obtained from
subjects who were dosed with a single dose of an investigational
compound. Plasma samples were collected over a 72 hour period. The
data show that the drug has a biexponential decline with a distinct
distribution phase and a slow elimination phase. I got a good fit when
modeling these data using zero order absorption with no lag time and
two comparmental model in WinNonlin. However, the %CV for the
estimates were very large. Any idea on techniques I can try to lower
the %CV?
Thanks
Martin
[What were the sample times? Assay error large? Large CVs for all
parameters? ROA? Oral? Zero order absorption, when did you stop the
'infusion' (fixed-adjustable)? How many data points versus number of
parameters - db]
Back to the Top
Dear Martin:
Probably the best answer I can suggest is to weight your data
by the Fisher information of your assay data, add a parameter for the
remaining environmental noise. Examine your data for the most
appropriate structural model. and after that, don't mess with reality.
As far as I know, after that there are no "techniques" to lower the
variability. You simply want to know the variability of your assay, of
that in the environment, (the 2 comprise the intraindividual
variability) and of that in the population. You might look at, for the
assay error -
Jelliffe RW, Schumitzky A, Van Guilder M, Liu M, Hu L, Maire P, Gomis
P, Barbaut X, and Tahani B: Individualizing Drug Dosage Regimens:
Roles of Population Pharmacokinetic and Dynamic Models, Bayesian
Fitting, and Adaptive Control. Therapeutic Drug Monitoring, 15:
380-393, 1993.
Along these lines, you might want to use a method of
population modeling that is statistically consistent (has the
guarantee that studying more subjects actually does get you parameter
estimates that are closer to the truth, and that makes no assumptions
about the shape of the parameter distributions. You might look at -
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.
Jelliffe R: Estimation of Creatinine Clearance in Patients with
Unstable Renal Function, without a Urine Specimen. Am. J. Nephrology,
22: 3200-324, 2002.
The next thought is - what do you wish to do with your model?
Publish it in a paper? Submit it to the FDA? Or use the model to
develop maximally precise dosage regimens for the patients who must
receive the drug? You might consider how to develop such regimens
using Multiple Model Bayesian Adaptive control. THIS is where you can
use well known techniques to minimize patient variability in response
to a particular dosage regimen. You might look at -
Jelliffe R, Schumitzky A, Bayard D, Milman M, Van Guilder M, Wang X,
Jiang F, Barbaut X, and Maire P: Model-Based, Goal-Oriented,
Individualized Drug Therapy: Linkage of Population Modeling, New
"Multiple Model" Dosage Design, Bayesian Feedback, and Individualized
Target Goals. Clin. Pharmacokinet. 34: 57-77, 1998.
Jelliffe R, Bayard D, Milman M, Van Guilder M, and Schumitzky A:
Achieving Target Goals most Precisely using Nonparametric
Compartmental Models and "Multiple Model" Design of Dosage Regimens.
Therap. Drug Monit. 22: 346-353, 2000.
Bleyzac N, Souillet G, Magron P, Janoly A, Martin P, Bertrand Y,
Galambrun C, Dai Q, Maire P, Jelliffe R, and Aulagner G: Improved
clinical outcome of paediatric marrow recipients using a test dose and
Bayesian pharmacokinetic individualization of busulfan dosage
regimens. Bone Marrow Transplantation, 28: 743-751, 2001.
Martin P, Bleyzac N, Souillet G, Galambrun C, Bertrand Y, Maire P,
Jelliffe R, and Aulagner G: Relationship between CsA trough blood
concentration and severity of acute graft-versus-host disease after
paediatric stem cell transplantation from matched sibling or unrelated
donors. Bone Marrow Transplantation 32: 777-784, 2003.
Martin P, Bleyzac N, Souillet G, Galambrun C, Bertrand Y, Maire P,
Jelliffe R, and Aulagner G: Graft Versus Host Disease: Clinical and
Pharmacolgical Risk Factors for Acute Graft-versus Host Disease after
Paediatric Bone Marrow Transplantation from matched sibling or
Unrelated Donors. Bone Marrow Transplantation 32: 881-887, 2003.
Bayard D, and Jelliffe R: A Bayesian Approach to Tracking Patients
having Changing Pharmacokinetic Parameters. J. Pharmacokin.
Pharmacodyn. 31 (1): 75-107, 2004.
Macdonald I, Staatz C, Jelliffe R, and Thomson A: Evaluation and
Comparison of Simple Multiple Model, Richer Data Multiple Model, and
Sequential Interacting Multiple Model (IMM) Bayesian Analyses of
Gentamicin and Vancomycin Data Collected From Patients Undergoing
Cardiothoracic
Surgery. Ther. Drug Monit. 30:67-74, 2008.
All the best. Hope this helps,
Roger Jelliffe
Back to the Top
Hi Roger:
Thanks for your insight. I am getting back into modeling after a gap
of about 15 years. I suppose I did not give sufficient information the
first time around. The study was done in 6 healthy volunteers and
plasma samples were collected after the administration of a single
dose at predose, 0.25, 0.5, 1.0, 2, 4, 6, 8, 10, 12, 16, 24, 48 and 72
hours. In terms of the number of samples collected, I think we have
done a fairly good job. Since I am getting back into modeling I
decided to take one subject and model the plasma concentrations to get
estimates of parameters which can then be used to simulate dosage
regimens. We also have data from administering multiple doses of the
same compound. However, I wish to start off slow, model the single
dose data first to get my feet wet before going into the multiple dose
data.
The fit using the built-in function in WinNonlin for one subject was
fairly good. The problem was that the %CV of the estimates was huge
(greater than 1000). I am seeing a similar trend in other subjects
too. I tried changing weights to see if the %CV of the estimates got
any better but without much luck.
I hope I have given a little more information that may help others
help me.
Regards
Martin!
[Looks like plenty of samples, well spaced (when does the alpha/beta
break occur? Do you see it on semi-log plot?). Good assay variability?
ROA; IV infusion or oral? If oral you have extra parameters and the
number of early data points may be insufficient to determine ka(k0)
and alpha (k21/k12) accurately. A sensitivity/identifiability analysis
might help - db]
Back to the Top
The following message was posted to: PharmPK
Martin,
It may be that your drug is not one of the simple ones that can be
adequately modeled with a constant Ka. Ka is never constant, it is a
time-dependent coefficient. Absorption is affected by a series of
complex
phenomena, which can include pH-dependent solubility and permeability,
dissolution/precipitation/redissolution, gut wall metabolism, lumenal
degradation, and transporter effects. The practice of assuming
constant Ka
is just as wrong as assuming all drugs are adequately modeled with
one-compartment PK.
I continue to be amazed at the detailed statistical arguments that are
often
presented when the underlying mechanistic behavior for oral absorption
is
glossed over with a constant Ka. You can fit models with constant Ka
but you
end up pushing other numbers around to compensate (and getting
incorrect PK
values as a result - maybe with large CV's?).
Walt Woltosz
Chairman & CEO
Simulations Plus, Inc. (NASDAQ: SLP)
42505 10th Street West
Lancaster, CA 93534-7059
U.S.A.
http://www.simulations-plus.com
E-mail: walt.aaa.simulations-plus.com
Back to the Top
The following message was posted to: PharmPK
Dear Martin,
The high parameter CV is may be due to the method used for data
analysis. If I understand you correctly, you applied the 'Standard
Two-Stage' (STS) method for getting the mean and sd of the model
parameters. Please note that STS may perform poorly, even in the case
of rich data sets, as shown in our paper:
Proost JH, Eleveld DJ. Performance of an Iterative Two-Stage Bayesian
technique for population pharmacokinetic analysis of rich data sets.
Pharm Res 2006;23(12): 2748-2759 (Erratum in Pharm Res 2007; 24(8):
1599).
An appropriate population analysis, e.g. by NONMEM, ITSB (see above
reference), or nonparametric methods (see reply by Roger Jelliffe)
will probably solve your problem.
best regards,
Johannes H. Proost
Dept. of Pharmacokinetics and Drug Delivery
University Centre for Pharmacy
Antonius Deusinglaan 1
9713 AV Groningen, The Netherlands
Email: j.h.proost.-a-.rug.nl
Back to the Top
The following message was posted to: PharmPK
Dear Martin,
Was the dose administered orally or through IV infusion ?
If it was administered orally, then there isn't a built in function in
WINNONLIN to fit that model
and you will have to write your own user model, with rate or time of
infusion as an additional parameter
An example of which is given as PK3 in Johan and Weiner's textbook.
If the dose was administered through IV infusion and you are using
built in model.
One suggestion could be to change initial estimates and bounds.
Winnonlin has fairly robust minimization algorithms, it is unlikely
that you are in a
local minima. One possibility is that if the true parameter estimate
is just outside the upper bound. In that case you may get a good fit
but your CV% will go up.
By default Winnonlin choses upper bound which is 10 times the value of
initial estimate.
If you are putting your faith on Winnonlin generated initial estimates
and winonlin generated bounds, change it
to winnonlin generated estimates and user specified bounds. Specify
larger upper bounds.
In that case Winnonlin will do a grid search to get the best initial
estimate.
You can also try to give your own initial estimates too.
Hope it helps.
Varun Goel
PhD Candidate, Pharmacometrics
Experimental and Clinical Pharmacology
University of Minnesota
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 "High parameter CV" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)