Back to the Top
Dear group members,
I am working on a drug that is extensively metabolized in the liver.
Its main metabolite is uniquely formed via the CYP2D6 enzyme system.
Due to genetic polymorphism of CYP2D6 the clearance of the parent
compound is highly variable. In order to account for this
interindividual variability I want to introduce the metabolic ratio
(ie. concentration of the metabolite (at time t) / concentration of
the parent (at time t)) as a covariate on the clearance.
My questions: Is it legal to use the drug and its metabolite
themselves as their own marker for phenotyping. Has anyone experience
with such a problem and/or know some literature?
Best regards.
--
Andreas Lindauer
University of Bonn
Department of Clinical Pharmacy
An der Immenburg 4
D-53121 Bonn
Back to the Top
The following message was posted to: PharmPK
Andreas:
You raise a very interesting question. We are trying to develop a
novel systems approach to pharmacokinetic/pharmacodynamic modeling,
which is physiologically based and only uses data obtained from the
individual. This is possible today using imaging based PK
measurements, which could be greatly extended with data on
pharmacogenomic expression. Because the availability of tissue for
genotyping is limited, what you suggest may be an indirect method of
generating pharmacogenomic data. Do you have any RT-PCR data from
animal studies that validates the relation between pharmacogenomic
expression and blood ratios of the metabolites in the CYP2D6 system?
I would also be interested in any additional information you get off-
line.
--
Professor Walter Wolf, Ph.D. Distinguished Professor of
Pharmaceutical Sciences
Director, Pharmacokinetic Imaging Program
Department of Pharmaceutical Sciences, School of Pharmacy
Chair, Biomedical Imaging Science Initiative
University of Southern California 1985 Zonal Ave., Los Angeles, CA
90089-9121
E-Mail: wwolfw.-at-.usc.edu
http://www.usc.edu/research/initiatives/bisi/
http://www.usc.edu/schools/pharmacy/faculty_directory/detail.php?id=59
Back to the Top
Andreas,
It has been shown in literature that the ratio of drug metabolite/
drug plasma/ urine concentrations is a valid phenotypic measure of
enzyme activity. There have been good enough correlations of AUC
ratios of metabolite/drug with the clearance values. Mutliple time-
point plasma/urine metabolite to parent drug concentrations ratios
are preferred over single time point ratios.
Back to the Top
Satya,
Can you cite any references? Thank you.
Raju
Back to the Top
Andreas
That would be cheating. You can't use the dependent variable or any
function of the dependent variable as a predictor in a system. If
you did, you would have no way to make predictions of future
observations.
Peter Bonate
Back to the Top
Please correct me if I have interpreted the original question in a
wrong way. Below are few references wherein metabolic ratios have
been used to phenotype CYP2D6 enzyme.
1. Tramadol as a new probe for cytochrome P450 2D6 phenotyping: A
population study. Clinical pharmacology & therapeutics 2005, 77:458-67
2. Quantitative effect of CYP2D6 genotype and inhibitors on tamoxifen
metabolism:Implication for optimization of breast cancer treatment.
Clinical pharmacology & therapeutics 2006, 80: 61-74
Back to the Top
The following message was posted to: PharmPK
Peter, Andreas,
I can think of two situations when accounting for the observed
metabolic ratio is useful/legal/not-cheating:
1. If you are going to perform the trial with individual
concentration-based dose adjustment (or aim for concentration-based
dose adjustment as a clinical practice), and will have first-dose PK
profile (or even 1 value) before adjusting the dose, then the model
that includes observed metabolic ratio can be useful to predict the
next-dose profile more precisely.
2. Metabolic ratio can be used instead of the genotype data (provided
this is validated somehow) to derive a categorical covariate: poor/
extensive metabolizer. This new covariate can be included into the
model. The new model can be used either in situations where you have
genotype data available, or to better describe PK of each of the
subgroups. This is similar to a mixture model where the assignment to
a particular group is based on the observed data.
On the contrary, if observed data (metabolic ratio in this case) are
used to decrease the inter-individual variability of clearance and
get better objective function values, then I agree with Peter that
this is an inappropriate use of the observed data.
Leonid
--
Leonid Gibiansky, Ph.D.
President, QuantPharm LLC
web: www.quantpharm.com
e-mail: LGibiansky at quantpharm.com
tel: (301) 767 5566
Back to the Top
Thanks to all the responders for their valuable comments.
Especially I would like to thank Leonid who seems to have telepathic
abilities. What she mentioned is exactly what I was thinking about.
The substance I am working with is venlafaxine, an antidepressant
drug. There are at least five papers "validating" the correlation
between metabolic ratio and genotype for this drug. Unfortunately, no
genotyping was done in the present study. Anyway, in my opinion, the
metabolic ratio may provide important information in clinical
practise (TDM) when the individual genotype isn't known (for dose
adjustment as Leonid explained). Of course, one has to be aware of
the limitations of the metabolic ratio (any influences on elimination
beside the patient's genotype, eg. co-medication, hepatic impairment).
Of course, I never had the intention to introduce a "cheating"
covariate solely to get better OFV. Sorry for not providing
sufficient information in my first mail.
--
Andreas Lindauer
University of Bonn
Department of Clinical Pharmacy
An der Immenburg 4
D-53121 Bonn
Back to the Top
I don't have the original question in front of me but I recall it had
to do with using the observed data as a covariate. If you have some
independent measure of metabolic activity then that is a valid
covariate but if you use the data on hand to help predict the data on
hand then that is not valid.
pete bonate
Back to the Top
The following message was posted to: PharmPK
Peter,
Just to remind the question: whether we can use observed data
(metabolic ratio) as a covariate in the PK model.
Now, assume that you have a latent variable (genotype in this case)
and was able to build a model of a probability for the patient to
belong to a particular type based on observed data (metabolic ratio
in this case). Then you can build a conditional model of PK based on
the estimated genotype and observed PK data. Sound OK so far? Now
assume that you can find the first model in the literature and
interested only in the second model. You can use the first model to
predict genotype, and then create the second (PK) model to describe
PK of each group. This is precisely the case under discussion. Looks
OK to me. This is a simplification but it should not introduce any
problems, especially if the separation of PK by genotype is strong.
Leonid
--
Leonid Gibiansky, Ph.D.
President, QuantPharm LLC
web: www.quantpharm.com
e-mail: LGibiansky at quantpharm.com
Back to the Top
The arguments seen kind of circular to me. If you are going to use
metabolite ratio to predict a concentration but you need the
concentration to get the metabolite ratio then why not take the
argument to the extreme and include your observed concentrations as
predictors in the model. Then you can get a perfect fit.
But I see your argument about if you were able to build a probability
model for genotype. I think that would be valid as long as you don't
use your observed data.
pete bonate
Back to the Top
The following message was posted to: PharmPK
Peter Bonate wrote:
> The arguments seen kind of circular to me. If you are going to use
> metabolite ratio to predict a concentration but you need the
> concentration to get the metabolite ratio then why not take the
> argument to the extreme and include your observed concentrations as
> predictors in the model. Then you can get a perfect fit.
I agree with Peter Bonate - using measured concs as a covariate to
predict the concs does not reveal anything new.
Leonid proposed the idea of using observed metabolic ratio to create
a categorical covariate. This wont give a perfect fit -- but only
because some information is discarded when converting MR to a
category. In principle it is the same flawed idea as using the
observed concs.
A mixture model to describe a multi-model distribution seems to me
the only way to help improve the description of the data without
'cheating'.
Nick
--
Nick Holford, Dept Pharmacology & Clinical Pharmacology
University of Auckland, 85 Park Rd, Private Bag 92019, Auckland, New
Zealand
n.holford.aaa.auckland.ac.nz
www.health.auckland.ac.nz/pharmacology/staff/nholford
Back to the Top
The following message was posted to: PharmPK
Nick,
You and Peter are answering a general question, without regards for
specifics. In general, I agree with you 100% that one should not use
data twice, as a covariate and as a dependent variable. In general,
mixture model could be a more correct way to deal with this type of
data.
Specific of this particular situation (I assume that this is true
based on the question) is that the metabolic ratio can separate the
overall population into two groups with high degree of accuracy. If
so, you can split the overall problem into two parts: separate
population into two groups first, and then use it as a covariate. The
only difference with the mixture model is that the mixture model fit
would run iteratively (separate population into the groups, describe
PK, separate again, describe PK, up to the convergence), rather than
sequentially. If two group in a mixture are widely separated, it
should not make any difference (iterative process would stop at the
first iteration). The next step is to identify the two population
with poor/extensive metabolizers. Again, I assume that this is
"validated" by multiple previous studies.
Using the two-step process, you may arrive at the simpler-to-use
model. For example, consider the situation with the concentration-
based dose adjustment: if you have first-dose data, metabolic ratio-
based test requires just a simple computation (of the metabolic
ratio) rather than the full nonmem-model run to identify the group
(genotype) for a particular patient.
Additional advantage (in this case) could be that the sequential
model (with metabolic ratio as a way to identify genotype) brings in
prior knowledge that the two groups can be separated by the metabolic
ratio, or even a particular threshold for this separation. It may
allow to build a better model ( with the help of the prior
knowledge ) than just a straightforward mixture model approach.
Leonid
--
Leonid Gibiansky, Ph.D.
President, QuantPharm LLC
web: www.quantpharm.com
e-mail: LGibiansky at quantpharm.com
Back to the Top
The following message was posted to: PharmPK
Hi
What we want to know is the concentration / dose for individuals at
steady state to be able to select dose. Should this be our phenotype?
In TDM [metabolite] adds value by helping diagnose non compliance.
Genotype is accepted as valid because it is easily defined, but it does
not perform well as a predictor of concentration.
Regards Matt
Dr Matt Doogue
Clinical Pharmacologist / Endocrinologist
Flinders Medical Centre, Flinders University
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 "Metabolic Ratio" as the subject | Support PharmPK by using the |
Copyright 1995-2011 David W. A. Bourne (david@boomer.org)